{"custom_id": "Rangamati_26163_64638_26676_65151.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JSZSVYN3ETXW57QS2B2tPd"}]}]}} {"custom_id": "Bogra_11799_12825_12312_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EDBqyAFwJ3dd1K5NkD8Au5"}]}]}} {"custom_id": "Sunamganj_16416_2052_16929_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Dwmiwdo1bDPvVh75cYwvGg"}]}]}} {"custom_id": "Mymensingh_12825_18468_13338_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1F3tsUwUyfq1w8rrbs4pgF"}]}]}} {"custom_id": "Lalmonirhat_9747_17442_10260_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hmy5kUoT2BAKzCZAMiQ1ZC"}]}]}} {"custom_id": "Natore_3591_17442_4104_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2yob4LnMHpB7SQoHjpTw3m"}]}]}} {"custom_id": "Patuakhali_13851_5643_14364_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CmafMDCg9daqQaBbQLbtJ9"}]}]}} {"custom_id": "Barisal_17442_18981_17955_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2SzAf8akSkEzUCR23Eq3Gc"}]}]}} {"custom_id": "Chittagong_24111_38475_24624_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DgTaX3pedi7C2t6t2sP7di"}]}]}} {"custom_id": "Habiganj_9234_30267_9747_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HXG7i1jwVtsP8vawoxEadG"}]}]}} {"custom_id": "Gazipur_19494_8721_20007_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WTLekwwYMRUhccTrpN2u5d"}]}]}} {"custom_id": "Rajshahi_18981_18981_19494_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MBvLQYqxs2869AAAFi9xMU"}]}]}} {"custom_id": "Sunamganj_28215_13851_28728_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8gT1P4DbeH2qdKz5WnotkS"}]}]}} {"custom_id": "Chandpur_1539_4104_2052_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Xh7pT3J3JyX4zQXVgAHJxF"}]}]}} {"custom_id": "Dinajpur_37449_34371_37962_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-E3C3QGx5X6rV1hFEX2p5AC"}]}]}} {"custom_id": "Madaripur_11799_15903_12312_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6txJEJFSwMEPdXLZC3bRF5"}]}]}} {"custom_id": "Brahamanbaria_513_27702_1026_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M5dTr9ztyXNxKSZEKMkwoG"}]}]}} {"custom_id": "Natore_16929_14877_17442_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-55QKtECr4hbvcLEXX9Y4bs"}]}]}} {"custom_id": "Magura_9234_3078_9747_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JsmMb4pkT3mHzZ68xD4sCA"}]}]}} {"custom_id": "Sylhet_7695_25650_8208_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BiKd46NwFiDNkrAapo1JmP"}]}]}} {"custom_id": "Pabna_14364_20520_14877_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EDpqzbSieFGWpDSsF9MEMK"}]}]}} {"custom_id": "Gaibandha_16929_27702_17442_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8GFwi1KjcZbGYsG9ja2MYX"}]}]}} {"custom_id": "Bandarban_11799_37449_12312_37962.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3pLt79eDx4NiNgq2rfdMRc"}]}]}} {"custom_id": "Satkhira_6156_15903_6669_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2ryTe1UDdtMmTQsFCKfTVd"}]}]}} {"custom_id": "Pabna_31293_24111_31806_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Pnu6uHYdzRvnDFrbogeXYJ"}]}]}} {"custom_id": "Netrakona_11799_26676_12312_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N7Ef3nrXmaTWpTKG4FE3i6"}]}]}} {"custom_id": "Netrakona_20007_17955_20520_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DpwtAfQ2C6TwCEeQapw6rA"}]}]}} {"custom_id": "Bandarban_18468_37962_18981_38475.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UyrCWMMz3NUdJ8mF9W2SU7"}]}]}} {"custom_id": "Gaibandha_16416_20007_16929_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WtzzHwv5v2wqVU56jUS6wJ"}]}]}} {"custom_id": "Bhola_16929_39501_17442_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8dUetHfXRVTP8BZkuT2HzH"}]}]}} {"custom_id": "Chuadanga_10773_7695_11286_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RkVKGH7UcnG2HWq1qxYrf4"}]}]}} {"custom_id": "Magura_7695_5130_8208_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9F4FmkySjAdShHXj7LccNv"}]}]}} {"custom_id": "Satkhira_16416_63099_16929_63612.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6913SczZgvHzYXp2G6VnHf"}]}]}} {"custom_id": "Nawabganj_22572_2052_23085_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3o5kiECAumXuAfjLDX6MNa"}]}]}} {"custom_id": "Brahamanbaria_15390_5130_15903_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SUcJ4fT7sLa8XVSq4Q71ki"}]}]}} {"custom_id": "Gazipur_7695_13851_8208_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3tnf4XbN3UQ12nhqVy9zwx"}]}]}} {"custom_id": "Nilphamari_10773_12825_11286_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TLk5nyqsbsfGq9FvKmAuf5"}]}]}} {"custom_id": "Habiganj_10260_15390_10773_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1iUwDjU7JcrhS6V2uqqHCv"}]}]}} {"custom_id": "Gopalganj_6669_16416_7182_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MgGPRq54kVVdiAVokpqtXc"}]}]}} {"custom_id": "Satkhira_17955_59508_18468_60021.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P2rAs8FVQVjbMVqoM9Ucfd"}]}]}} {"custom_id": "Sunamganj_35397_6669_35910_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8aJrP8ds8ut1fMrUAMGtTv"}]}]}} {"custom_id": "Rangamati_7695_56943_8208_57456.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CitVfbshWT98o6XYfhEtzH"}]}]}} {"custom_id": "Khagrachhari_13338_24624_13851_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LNJrsK3fQ8SYqGaKzfxjoY"}]}]}} {"custom_id": "Noakhali_9747_29241_10260_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DCiMyRBcT6ts1mtdSXVAv2"}]}]}} {"custom_id": "Barisal_21546_16416_22059_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NQuWzoM5Ra8jb6uge22ayt"}]}]}} {"custom_id": "Rangamati_15903_34371_16416_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MvUurtM85VXJAd6CAhPmWY"}]}]}} {"custom_id": "Comilla_9234_7695_9747_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Df9od9m6G3MC6APjbmHiGd"}]}]}} {"custom_id": "Jhenaidah_16416_2565_16929_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KpWgByM5EdirGR26B7WDuT"}]}]}} {"custom_id": "Rajshahi_3078_7182_3591_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LX8buFZLpjkayPybjtfRra"}]}]}} {"custom_id": "Bogra_16929_14364_17442_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RjdUbTuDivjpLnKLq6j2sD"}]}]}} {"custom_id": "Chittagong_32832_52326_33345_52839.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GbtZV2YdgMVxnnWiYoa6nj"}]}]}} {"custom_id": "Sirajganj_20007_1539_20520_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VJLHnizmEfu17KiPMJoGLV"}]}]}} {"custom_id": "Shariatpur_1539_8208_2052_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BzmF8AKgihRh8AZ1WUz9sV"}]}]}} {"custom_id": "Comilla_30267_32319_30780_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XBtAueGURJAXgyveaqVboT"}]}]}} {"custom_id": "CoxsBazar_17955_47709_18468_48222.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vhq6B6T2q5BWJG7XfhoUtX"}]}]}} {"custom_id": "Noakhali_16416_18468_16929_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BZikgMJtFdfJQrwrNfp5Eg"}]}]}} {"custom_id": "Habiganj_4104_18981_4617_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2MzFu7ymbZicKGer7QHu1Z"}]}]}} {"custom_id": "CoxsBazar_7182_5130_7695_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Xuydy1bpCTYbj5o5SjzPQR"}]}]}} {"custom_id": "Nilphamari_5130_17442_5643_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QFxvoXj8fBYqWGndwoXZsr"}]}]}} {"custom_id": "Mymensingh_19494_20520_20007_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7D6gQFiigWZeCv9syvdPc7"}]}]}} {"custom_id": "Meherpur_6669_6669_7182_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Eh5b7yoiisg6Wj9Svas7Dk"}]}]}} {"custom_id": "Jessore_22059_11286_22572_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NfHtHbyebsDpcTf1ZnSALQ"}]}]}} {"custom_id": "Noakhali_11799_38988_12312_39501.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XmDi3hbCJLJ4A4tefK3cUK"}]}]}} {"custom_id": "Thakurgaon_13851_24111_14364_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5ezg48xqxqNcn65uC8DuEC"}]}]}} {"custom_id": "Kurigram_12825_25137_13338_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RDRjbKkJVkALoBmZVh8BJm"}]}]}} {"custom_id": "Chittagong_25650_17442_26163_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QHzz3twJqQavhrYJx3Q8ht"}]}]}} {"custom_id": "Feni_1539_17442_2052_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-593pMfPJzBMYztnshE92HL"}]}]}} {"custom_id": "Bagerhat_13851_25650_14364_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3dEDJ14oEA48K1BgRtMdya"}]}]}} {"custom_id": "Khagrachhari_7695_52326_8208_52839.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WVSbmbockV4jrF4974qM8T"}]}]}} {"custom_id": "Joypurhat_13338_10260_13851_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J44JFQQMafbHTygwvbbv1M"}]}]}} {"custom_id": "Sylhet_15903_23598_16416_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4Ed3nY8AzCSQ12vPEzaXnU"}]}]}} {"custom_id": "Lakshmipur_11799_26676_12312_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NmUei8J3U1nrNwsigPb594"}]}]}} {"custom_id": "Bhola_6669_1539_7182_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3pvW1LonRnihHdMs6iQQbR"}]}]}} {"custom_id": "Barisal_25650_15903_26163_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DpfxrpQQuhoWrNkwKt18nx"}]}]}} {"custom_id": "Rangpur_23598_16416_24111_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AtmwcXUYBXiseJDVmEhbGa"}]}]}} {"custom_id": "Noakhali_17442_26163_17955_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-75t2wbPYoVq9XuPhzJE7uH"}]}]}} {"custom_id": "Sunamganj_12312_4104_12825_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JGS7RJMfZRg9kgoTyfHTnd"}]}]}} {"custom_id": "CoxsBazar_4104_21546_4617_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DuaQBDw2ywHztXYn74okmu"}]}]}} {"custom_id": "Jhenaidah_17955_16929_18468_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4oMgbYdcp5ut3nGzKWYApG"}]}]}} {"custom_id": "Bogra_25137_6156_25650_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ATG7boBoo1U9Lf2Rx6ZM9W"}]}]}} {"custom_id": "Chandpur_6156_8208_6669_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5JpegJ2HJmezFojf7cSjdv"}]}]}} {"custom_id": "Shariatpur_13851_11286_14364_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7mavk4gsDZa8LYhfTtLEgS"}]}]}} {"custom_id": "Khulna_11799_43605_12312_44118.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YC1wvxF6R2PgRpwtzABJDM"}]}]}} {"custom_id": "Chandpur_5643_1026_6156_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Prxnu54rBcrEqR2HbPDHVp"}]}]}} {"custom_id": "Faridpur_21546_5643_22059_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PDZgP147c7Y8zTFtpm6znA"}]}]}} {"custom_id": "Barisal_20520_4104_21033_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3DBwXd1N48rQDwYF75g7z9"}]}]}} {"custom_id": "Gaibandha_3591_24111_4104_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K1ht6EZwTbAg5iDEw4XqL1"}]}]}} {"custom_id": "Sylhet_21033_6156_21546_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Er9zyATyBAcTm17n3EJXEL"}]}]}} {"custom_id": "Sylhet_29754_11286_30267_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GEmorSkyvwaSnyEqkFwSRr"}]}]}} {"custom_id": "Patuakhali_11799_11799_12312_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ETxfrCRzpS55oyqUqP9zrh"}]}]}} {"custom_id": "Shariatpur_10260_6156_10773_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MEWiYHhRWkAM8LccPp9W6r"}]}]}} {"custom_id": "Rangamati_15390_35910_15903_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4Ka42g7XoKf58rw7zfYbQg"}]}]}} {"custom_id": "Bhola_8208_40014_8721_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6oThKwTQXXANbP6omrjJZb"}]}]}} {"custom_id": "Rangpur_11286_24624_11799_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9uEWuHHDxuAUQBYD4cRJxb"}]}]}} {"custom_id": "Satkhira_18468_11286_18981_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4epnGSsBbxLDafrMVJShw6"}]}]}} {"custom_id": "Thakurgaon_21546_6669_22059_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NhWjRXGayUNYhmCbkkCtb7"}]}]}} {"custom_id": "Gopalganj_5130_9747_5643_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9n1SszikZjFPdoLy361bW6"}]}]}} {"custom_id": "Bogra_19494_24624_20007_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2AFnNMH5AUnVU8f7kMdkiN"}]}]}} {"custom_id": "Tangail_8721_9747_9234_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D7r73BfhrhQQrcRq43gNvn"}]}]}} {"custom_id": "Chandpur_17442_15390_17955_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5Xt9t7i4za4sbc7H5up5dS"}]}]}} {"custom_id": "Khagrachhari_18468_14364_18981_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QcwWRhLt8WGdRdaNFBY5yp"}]}]}} {"custom_id": "Satkhira_6669_52326_7182_52839.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MREB1amnpEcev1rEvGsTCp"}]}]}} {"custom_id": "Pirojpur_6669_20520_7182_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CouCguisXeA9viCgm74fBc"}]}]}} {"custom_id": "Sherpur_16929_5130_17442_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-USucFZaUobdvuMNHpWC2Aj"}]}]}} {"custom_id": "Kishoreganj_12312_7182_12825_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GhDokyyFMEVCNUk5yGqs8g"}]}]}} {"custom_id": "Kishoreganj_28215_13338_28728_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DM3diG6xWERKGsvX3ifWnU"}]}]}} {"custom_id": "Patuakhali_20007_3591_20520_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NYoCGcYhSSed7GokckbcKg"}]}]}} {"custom_id": "Gopalganj_12825_16416_13338_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3f2AEDj8ZeYzPSdr75hNQA"}]}]}} {"custom_id": "Jessore_14364_21033_14877_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-G3EAr3SN1665SJgrd4KmPM"}]}]}} {"custom_id": "Rangamati_13851_40527_14364_41040.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8QgpU73dcFJ4v9St77KcPV"}]}]}} {"custom_id": "Bhola_5130_6156_5643_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LyWPWGnPdXh7XHDhFSbHHd"}]}]}} {"custom_id": "Bandarban_15390_23085_15903_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2kUYM48GVuBA4FLae5daCX"}]}]}} {"custom_id": "Tangail_15903_4104_16416_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PWvzeuy5sj16f4965UM1et"}]}]}} {"custom_id": "Gazipur_6669_2565_7182_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QEwCzxTnm8Deo84KSRjnK2"}]}]}} {"custom_id": "Khagrachhari_8208_52326_8721_52839.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QtjixTREQSRZpuVX64ytjR"}]}]}} {"custom_id": "Bandarban_13851_40527_14364_41040.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Dv6UghE9d3TR9bsPjTgTMJ"}]}]}} {"custom_id": "Sirajganj_13851_13338_14364_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ALua9J88rEaEbBzBxRjzye"}]}]}} {"custom_id": "Habiganj_15390_17955_15903_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3FRXi8gKbEUpMsGSGiuvUH"}]}]}} {"custom_id": "Satkhira_9234_38988_9747_39501.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LWfurS9m5TRyPQc22hpy45"}]}]}} {"custom_id": "Chittagong_29754_30780_30267_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-279XgCFjEFGMhNpgfZQPpm"}]}]}} {"custom_id": "Brahamanbaria_23598_16929_24111_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GeuQ2ysrC5bSvdYqQoNmiZ"}]}]}} {"custom_id": "Habiganj_16929_23085_17442_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9amBd4B6e7YB12UTHfDpd1"}]}]}} {"custom_id": "CoxsBazar_10260_2565_10773_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4Zeoy5xtpogTTiUgX8D38t"}]}]}} {"custom_id": "Barguna_12825_5130_13338_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UPEDCqYgYUN7CtC5A1pwSL"}]}]}} {"custom_id": "Kishoreganj_10773_7695_11286_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PXMf1u9ZFvep9oBEutP3Js"}]}]}} {"custom_id": "Kishoreganj_23085_8208_23598_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6L9gtwjonGuwPt3RLyCC6b"}]}]}} {"custom_id": "Bandarban_5643_38475_6156_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2r8f1hJx6WYZ9R9ot9Hedz"}]}]}} {"custom_id": "Habiganj_5130_12825_5643_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1dRUwT9m7S52MbjaPmPXa3"}]}]}} {"custom_id": "Bandarban_25137_49761_25650_50274.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LzGfdE9vuKy5vQW8NpQi3H"}]}]}} {"custom_id": "Sylhet_22572_21033_23085_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U1L99Uo8HP4gsa1XgGe6fo"}]}]}} {"custom_id": "Noakhali_13851_42066_14364_42579.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7gGmTA9i6sSpbDL3Qr6myv"}]}]}} {"custom_id": "Mymensingh_7182_18468_7695_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8YhcHftx8zahE7rv7gvF9G"}]}]}} {"custom_id": "Faridpur_6156_12825_6669_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X8N3Siu9MFVSN99DT8p66P"}]}]}} {"custom_id": "Maulvibazar_6669_27702_7182_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Wdq5ZGWVVFBh8RyeFvHzU9"}]}]}} {"custom_id": "Naogaon_5130_22572_5643_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9EUKvoz46ydshcJiBPVhEP"}]}]}} {"custom_id": "Jhalokati_8208_3591_8721_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K92DprwggK4sji3RkN7ehg"}]}]}} {"custom_id": "Sylhet_10773_11799_11286_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YJwo3Ehcg7LVQUGZ9HyCb3"}]}]}} {"custom_id": "Pabna_7182_8208_7695_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N8gtaDzG5isDoQChnjwKex"}]}]}} {"custom_id": "Panchagarh_18981_22572_19494_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-99frQxEgNuWGrsiSqgvqfD"}]}]}} {"custom_id": "Rajbari_7695_4617_8208_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RzSzM26k3wGLjA5dmSjEfN"}]}]}} {"custom_id": "Nilphamari_2052_12825_2565_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4Zfvd4miG7ZD3iwJpNegHB"}]}]}} {"custom_id": "Patuakhali_19494_3591_20007_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AJu5pYCiivUbEf3yK3Rdqx"}]}]}} {"custom_id": "Jamalpur_22059_28215_22572_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7WxpRAEwpZcKHL38RnXnFD"}]}]}} {"custom_id": "Sylhet_6156_14364_6669_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KjGa9PtgfS3DeQrkCpcHwg"}]}]}} {"custom_id": "Rangamati_25650_53865_26163_54378.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DstupRe6Shi9MLXwX1cYYj"}]}]}} {"custom_id": "Rajbari_18468_12825_18981_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RijNPubQXHgNWjDYQaLdga"}]}]}} {"custom_id": "Kurigram_6669_28728_7182_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F2v8SAwpDnYRonRrUtwoAK"}]}]}} {"custom_id": "Patuakhali_4617_40014_5130_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7nPb1WS3yMRVQkSwSn6FF3"}]}]}} {"custom_id": "Nawabganj_8208_10260_8721_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YLfJoAtEoQnCTT8xktHZhT"}]}]}} {"custom_id": "Patuakhali_4104_15903_4617_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3CX3H1z323z3uctE5X1RGq"}]}]}} {"custom_id": "Gaibandha_9747_18468_10260_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M9WVk1VtKDwSTnKzUFQfz5"}]}]}} {"custom_id": "Bagerhat_8721_46683_9234_47196.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bbo9mig67x1nMYfQmVSq3x"}]}]}} {"custom_id": "Pabna_30267_22059_30780_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-14AhXbMnf7vcREvKSE1ysL"}]}]}} {"custom_id": "Pabna_29241_13851_29754_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XYVLTX2SM98rumU9iJMk1G"}]}]}} {"custom_id": "Naogaon_24111_19494_24624_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KmkgpaqMJQtdrXshttNiZc"}]}]}} {"custom_id": "Satkhira_6669_31806_7182_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DwTz9s1ZFgBD3LhKsuEzhk"}]}]}} {"custom_id": "Kurigram_11799_20520_12312_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7aYn1fKMWXugjFh7GHRKbv"}]}]}} {"custom_id": "Mymensingh_12825_9234_13338_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Eq7kossFsGVxgXnGAsQpoT"}]}]}} {"custom_id": "Kishoreganj_14364_21033_14877_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8rQ8JRmPLYzejiLFJy2Hvq"}]}]}} {"custom_id": "Nawabganj_19494_2565_20007_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PEt3KCJAqreb1DoYX66m79"}]}]}} {"custom_id": "Rajshahi_27702_14877_28215_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JgoLn1CqwoPBj52AYGKweN"}]}]}} {"custom_id": "Naogaon_16416_13851_16929_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7wVEW5V1cKqo6E3PP1tJWt"}]}]}} {"custom_id": "Narsingdi_6156_12825_6669_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PYAE2GxtvEcLARRjsCztki"}]}]}} {"custom_id": "Satkhira_7182_42066_7695_42579.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9JiZHsaLL1no41SfgoCu5H"}]}]}} {"custom_id": "Khulna_2565_33345_3078_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6jMbF4UMnwTYTiPHa45mBU"}]}]}} {"custom_id": "Jamalpur_8208_24111_8721_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-21NoARTGyHJyEkHFMSRBfn"}]}]}} {"custom_id": "Chittagong_39501_52839_40014_53352.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3wpngw2FbPuuZMCv8D9qme"}]}]}} {"custom_id": "Rangamati_19494_45657_20007_46170.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WzPpnCz443ZbisStVByhJ1"}]}]}} {"custom_id": "Tangail_9747_2565_10260_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-B5AL5n7NpjayiKze68iAHK"}]}]}} {"custom_id": "Faridpur_25650_10260_26163_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7EqopRbtzJif5priyehdka"}]}]}} {"custom_id": "Satkhira_5643_11286_6156_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WHBUFoGe2fC2kSAZhSWTRF"}]}]}} {"custom_id": "Gaibandha_25137_12825_25650_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5rAhn4ta7fXfX7gpVegmFC"}]}]}} {"custom_id": "Jessore_14364_25650_14877_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SfBKqKuedceBdgvVa3aAr2"}]}]}} {"custom_id": "Jhenaidah_18981_9747_19494_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9TyTDEQc5wa627aiX29nSn"}]}]}} {"custom_id": "Khagrachhari_17955_6669_18468_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8H5EcTv43PpHv7WqvoJC9q"}]}]}} {"custom_id": "Mymensingh_24111_22059_24624_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-H3eDFXZMhXB4qKcg87LMYB"}]}]}} {"custom_id": "Jhalokati_11286_7695_11799_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RG2Bxd2y8PqUPBZxXJkUPx"}]}]}} {"custom_id": "Madaripur_0_17955_513_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6JU114CzmiTZwHgZBHDzyW"}]}]}} {"custom_id": "Barisal_18981_4104_19494_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4hWHqMJo7gK6EaTX7P6n8n"}]}]}} {"custom_id": "Nawabganj_16929_8208_17442_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4WZvvMTcgBejUznzcqGvMC"}]}]}} {"custom_id": "Khulna_11286_31293_11799_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RDTHBVbzkoTgGcAKYGQeof"}]}]}} {"custom_id": "Khagrachhari_16929_25137_17442_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2GrDG3iPsid1jRuLPVTFKw"}]}]}} {"custom_id": "Kurigram_9747_34371_10260_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FzfAyL6XQ8P4S21Q4vnAjC"}]}]}} {"custom_id": "Faridpur_14877_13851_15390_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EDjJF8PsGPEnDX4DHGDLNG"}]}]}} {"custom_id": "Sherpur_6669_7695_7182_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Kmbh6KzoHoPkguuVDZSriL"}]}]}} {"custom_id": "Kishoreganj_5643_17955_6156_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NWiKVGrjW8tMqzsifHKgFE"}]}]}} {"custom_id": "Kurigram_17955_13851_18468_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5gnwVH3yMv2N6KEQmYBj2Q"}]}]}} {"custom_id": "Satkhira_6156_11286_6669_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EHBKFC5YpSLSWMEKnutopG"}]}]}} {"custom_id": "Kushtia_16416_12825_16929_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8PnjxomYv7qUHxdfDSKZra"}]}]}} {"custom_id": "Barisal_25650_7695_26163_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9RCrKYWas11gxYPzJazH1f"}]}]}} {"custom_id": "Satkhira_14877_37962_15390_38475.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5nw86rfBZDpgKSKFDztrnY"}]}]}} {"custom_id": "Gaibandha_9234_15903_9747_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5symooc1tkfNFB3EKEqzbe"}]}]}} {"custom_id": "Habiganj_7182_27189_7695_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2CQWkZTjxEkYek9sa1yoLt"}]}]}} {"custom_id": "Brahamanbaria_24624_22059_25137_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UpUfzC8eWc8Wz3QaVdX3pu"}]}]}} {"custom_id": "Nawabganj_13851_16416_14364_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JHLVGvXuCBVMHwaxZqb2kD"}]}]}} {"custom_id": "Naogaon_12825_20007_13338_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YDG6H5uZ6YcUyfRLjjSi3R"}]}]}} {"custom_id": "Dinajpur_18468_16929_18981_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BZvqqdVoV6q9vtLgcWkAXU"}]}]}} {"custom_id": "Mymensingh_20520_40014_21033_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L545fZnswv49ZE4Hkz7ECU"}]}]}} {"custom_id": "Rangamati_30267_90288_30780_90801.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Si9FTZYpAShJawMRi9FnaZ"}]}]}} {"custom_id": "Rajbari_13338_15903_13851_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-73Ui6eSe4TxAb12rQMFgPu"}]}]}} {"custom_id": "Comilla_16929_6156_17442_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ro9bo62q55yqPb7AMovVr7"}]}]}} {"custom_id": "Bogra_22059_12312_22572_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YQEBrH4GqcfEEYtv1xoKR8"}]}]}} {"custom_id": "Khagrachhari_4104_22572_4617_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9vVFGtKyatcMXedHaRFM9q"}]}]}} {"custom_id": "Khulna_8208_32832_8721_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Uj2LP6Gr2TPXXsa1ra4bGL"}]}]}} {"custom_id": "Sylhet_5643_5130_6156_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Xd7fjiqpsL9o5jvKdGLQ5V"}]}]}} {"custom_id": "Noakhali_6669_5643_7182_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Jgvp9VgcudQUNRnCxPNJBi"}]}]}} {"custom_id": "Rangamati_15390_40014_15903_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6JBDPoaLj79LRwr897Wu83"}]}]}} {"custom_id": "Habiganj_10773_11286_11286_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Qb5RGyMt29HfwBSJYeYpje"}]}]}} {"custom_id": "Maulvibazar_16929_9747_17442_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HLMaNjg8hfCGY4oDW5RpE4"}]}]}} {"custom_id": "Sylhet_4104_7695_4617_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AiiRFwFFnstbpyBpScJEmE"}]}]}} {"custom_id": "Bagerhat_11286_44118_11799_44631.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3hVgE3E9bRJ8P9CnDkUHVL"}]}]}} {"custom_id": "Jhenaidah_17442_23085_17955_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-E94etHRitSD1kJkWd6kgr8"}]}]}} {"custom_id": "Naogaon_7182_13851_7695_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KKRVYZdHoofiSwVKJ4z4bd"}]}]}} {"custom_id": "Pirojpur_3591_10260_4104_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BfyQ86aFdgKD6FjYzrpGqf"}]}]}} {"custom_id": "Sylhet_4104_9747_4617_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JgWYKAxr1rcyh29uD2gqz6"}]}]}} {"custom_id": "Netrakona_7182_4104_7695_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4ADSDgpodDCoUjz7xPaw9U"}]}]}} {"custom_id": "Dinajpur_8721_10260_9234_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8ogpgCDXWatHmfLKCDJKkD"}]}]}} {"custom_id": "Jhenaidah_27189_10773_27702_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GxWhaAsiAStjVzeid6UYpG"}]}]}} {"custom_id": "Gopalganj_5130_16416_5643_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UYiyAEs3LTWCL2p5LhHtvs"}]}]}} {"custom_id": "Jessore_17955_9234_18468_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-23q3RgxjyoKg6hC9mHpiFm"}]}]}} {"custom_id": "Chittagong_38988_41040_39501_41553.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2pZbwPgm1erQVVKwShEH9V"}]}]}} {"custom_id": "Gazipur_8208_13338_8721_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-E4gXFy2E9noVvkfZfujS5L"}]}]}} {"custom_id": "Maulvibazar_9234_19494_9747_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Jf1GxGhxLceXjWpbmiRteC"}]}]}} {"custom_id": "Chittagong_22572_33858_23085_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VXjVn7pMDoQVhZVzD9p1CZ"}]}]}} {"custom_id": "Faridpur_8721_17442_9234_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LGUtrnRiiZD7FygZWtcZsV"}]}]}} {"custom_id": "Sylhet_38475_14364_38988_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R4jBedwAfwhLJHc1D3nu3n"}]}]}} {"custom_id": "Kushtia_22572_16416_23085_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GLScQQ4XcchJQvSE39F2qw"}]}]}} {"custom_id": "Jhalokati_14877_6669_15390_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2KJZPn1XZz2NN53Nz4CUbQ"}]}]}} {"custom_id": "Tangail_13338_10773_13851_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DrueigTnCVLnu2D3AAeyeD"}]}]}} {"custom_id": "Munshiganj_17955_11799_18468_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MrBgYFGDgSQgV1QQ72t324"}]}]}} {"custom_id": "Khulna_15903_44631_16416_45144.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gt2Jc64KBauAPvQ4vbRrqq"}]}]}} {"custom_id": "Rajshahi_18981_9234_19494_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UoyLF7GcPXWWkkgRhX4VK6"}]}]}} {"custom_id": "Bagerhat_8721_53352_9234_53865.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VQUbMWBubJgLFf8fP5yydi"}]}]}} {"custom_id": "Barguna_11799_11799_12312_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-823RGtjLPdc9TpWR3uy9Sd"}]}]}} {"custom_id": "Gopalganj_18468_16929_18981_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Kjtbkczyo5K8Lxkx8U8GLJ"}]}]}} {"custom_id": "Khulna_5130_24624_5643_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JjMxKzbeJ7VUvwj5ttJ3jX"}]}]}} {"custom_id": "Patuakhali_13851_23598_14364_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AayUxTuyu9FNAgM2Uugsut"}]}]}} {"custom_id": "Bagerhat_10260_44118_10773_44631.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6Qo1RJKpyr1iuuFaNbaiiU"}]}]}} {"custom_id": "Pirojpur_7695_26163_8208_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AUPU4gwcsT6bGPGxnrSAvF"}]}]}} {"custom_id": "Mymensingh_7695_17442_8208_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YC6tdF9cEJaJh46ccHp7Jn"}]}]}} {"custom_id": "Sylhet_33345_6669_33858_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WX2d94KbDqHGjv9R11Rfe8"}]}]}} {"custom_id": "Nilphamari_9234_26676_9747_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KGRKtRHpJxPdm581z8GoPb"}]}]}} {"custom_id": "Bhola_6669_7695_7182_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9ghxNaudhge7hEE3i7525Z"}]}]}} {"custom_id": "Rajshahi_1026_13338_1539_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MebLrmD1BbXwF6Vf4yHKpr"}]}]}} {"custom_id": "Munshiganj_7182_9234_7695_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Db7q6hk6XcBEKqgsbvTuxa"}]}]}} {"custom_id": "Bhola_5130_9234_5643_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9x1ysNa8eH9hyFgoYQ96Am"}]}]}} {"custom_id": "Noakhali_11286_43092_11799_43605.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V14DMDeLgU3d4kriPM28R5"}]}]}} {"custom_id": "Nilphamari_5130_13338_5643_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QD7pAgutBGT4H9pnM6x2B9"}]}]}} {"custom_id": "Sylhet_12825_8721_13338_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XLHFDL7esbqXkgVq4LUqzq"}]}]}} {"custom_id": "Comilla_16416_25137_16929_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NN8miPnu3J3f5VzwnpVtTC"}]}]}} {"custom_id": "Jessore_21546_9747_22059_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LvvWPjj4Qyzxr53pDgJiEZ"}]}]}} {"custom_id": "Nawabganj_13338_5130_13851_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3n7DDLRo4Hi2X3dieLXiMA"}]}]}} {"custom_id": "Comilla_27189_25137_27702_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CJLKxPZ8Z3tJFGqXNoNLAa"}]}]}} {"custom_id": "Dinajpur_29241_36423_29754_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ab1JotxWFmiXgUExz6sThP"}]}]}} {"custom_id": "Maulvibazar_28215_11286_28728_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GcLFHHZgrTbXSMPypt5GUf"}]}]}} {"custom_id": "Feni_7182_25650_7695_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7nV7gVjuFBoku9b2DtMDhd"}]}]}} {"custom_id": "Rangamati_18468_62073_18981_62586.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-148p1rjLiAqea5ptu4daJs"}]}]}} {"custom_id": "CoxsBazar_22059_51813_22572_52326.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HMPi7RsvbsaRwV4XGChB16"}]}]}} {"custom_id": "Jamalpur_1539_14364_2052_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3etU9KYdRxHRyMSXqaQ7yv"}]}]}} {"custom_id": "Barisal_22059_23085_22572_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M3oMxzwKPN7jehdYMwwEtg"}]}]}} {"custom_id": "Jessore_30267_20520_30780_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7aCKTyc7MCvXNBKn5NmysJ"}]}]}} {"custom_id": "Thakurgaon_12312_13338_12825_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JvoPh9t6SXCizmVQKhBfSy"}]}]}} {"custom_id": "Comilla_24624_9234_25137_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U8w3rEVhSgKTAFiaLon3q3"}]}]}} {"custom_id": "Sunamganj_27189_22059_27702_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4ubyPKXcsLFK16a59ZY2Ch"}]}]}} {"custom_id": "Chittagong_34884_28728_35397_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GvbLvPRbxvCAAUH6Ven1on"}]}]}} {"custom_id": "Khagrachhari_17442_22572_17955_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Pek8f98EcpNE3UZR5rbBnz"}]}]}} {"custom_id": "Rajbari_10773_7182_11286_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JRo1bQCuWNnoiTDW66r73R"}]}]}} {"custom_id": "Rajshahi_10773_4104_11286_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9uKkhZcqsdswMdEcbFzGXs"}]}]}} {"custom_id": "Mymensingh_23598_24111_24111_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P3p36qpVipMSvmwV8GLWu5"}]}]}} {"custom_id": "Gopalganj_4617_7695_5130_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-212RUMjcEbx5ZV7xswuzoJ"}]}]}} {"custom_id": "Bagerhat_6669_20520_7182_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ES2C2pd6yRCktEFyvHxYnu"}]}]}} {"custom_id": "Nilphamari_10260_27189_10773_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3xyac6dDtwkcuLA1K5ZmDt"}]}]}} {"custom_id": "Tangail_20520_32319_21033_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DFMd9sHgwGjPYQbrz9JkqN"}]}]}} {"custom_id": "Lalmonirhat_23598_27702_24111_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Jy8gV7ux8W7zPdHfK88XQB"}]}]}} {"custom_id": "Rajshahi_26676_18468_27189_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FF6vvYJZmg8nbMM8h2LWia"}]}]}} {"custom_id": "Habiganj_18468_28215_18981_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VSDBUBDieuLoLsYKpSqUHJ"}]}]}} {"custom_id": "Maulvibazar_20520_20007_21033_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9Ab5XyCq4SQj1DFM7hXfrY"}]}]}} {"custom_id": "Bogra_25137_23085_25650_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S4gfdWV3wTAPE5f1j9b6z6"}]}]}} {"custom_id": "Chittagong_33858_36936_34371_37449.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RAZxqENHBu9CDNZ6orjzp7"}]}]}} {"custom_id": "Faridpur_7695_6669_8208_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ntk5S6kBA6CQwcneAedUfd"}]}]}} {"custom_id": "Bandarban_23085_25137_23598_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LPXMvPfEmpru5SNTg7n6WQ"}]}]}} {"custom_id": "Bagerhat_15903_17442_16416_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RQ8HmNi2sft9yenZecPvGd"}]}]}} {"custom_id": "Narayanganj_13851_6156_14364_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Jw8eQ1GVb7L2e8oS2WZyUw"}]}]}} {"custom_id": "Natore_0_18468_513_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ErUEMFfyT8i3M5BgCCz6Um"}]}]}} {"custom_id": "Bagerhat_12312_17442_12825_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Xu5wxy5fYmFSv9yxtnfA66"}]}]}} {"custom_id": "Chittagong_31293_49761_31806_50274.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2JSWMcgbHpWHFckz8TCkB6"}]}]}} {"custom_id": "Jessore_6156_6669_6669_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M4WTWS3T5ZEBWVg8EV2QyD"}]}]}} {"custom_id": "Habiganj_10773_25650_11286_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WsMXYHJj6MttG19hnLA5ho"}]}]}} {"custom_id": "Kushtia_3591_7182_4104_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1iE5PjqSbRDPPDkwKijudE"}]}]}} {"custom_id": "Chittagong_22059_10773_22572_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2SNZdTa219B2bVhZivMSCD"}]}]}} {"custom_id": "Bagerhat_6669_39501_7182_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Y8f9kPNG56N4sFa1gXDTrm"}]}]}} {"custom_id": "Jhenaidah_24624_6156_25137_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UYP3fh2pv84RA3KCPagZRY"}]}]}} {"custom_id": "Jhenaidah_23085_16929_23598_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QKP7FHqeKaGmCtmBD6Sb6S"}]}]}} {"custom_id": "Chittagong_19494_6156_20007_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FiDVSxidafL7dwt9RG3JG3"}]}]}} {"custom_id": "Kishoreganj_20520_26676_21033_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XxGPGMrtZ9v45JCThLZPMU"}]}]}} {"custom_id": "Jessore_24624_12825_25137_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JwgpMi4hYkAmSieMNtEwAB"}]}]}} {"custom_id": "Rangamati_20520_61560_21033_62073.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CvphNY5yRi2csNKL1Qm7jo"}]}]}} {"custom_id": "Jhenaidah_27189_15390_27702_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HEy19Pf75uU5xFMRAVCxLV"}]}]}} {"custom_id": "Tangail_21546_13851_22059_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QCTrrK7EzFJFSgdmaRebvu"}]}]}} {"custom_id": "Sylhet_9747_9234_10260_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Pw7yxxzFV6n91gysshzRje"}]}]}} {"custom_id": "Bagerhat_3078_41553_3591_42066.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1sHe5HPRzoFfdTh9UVFcqA"}]}]}} {"custom_id": "Barguna_8721_28728_9234_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AQhLJWnnHZBXcixQ3cKAgq"}]}]}} {"custom_id": "Kurigram_13338_18981_13851_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9E86yNCZ7BdPvm4H6qdK3y"}]}]}} {"custom_id": "Bhola_7695_35910_8208_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GZMK98JHugjaHKL4UxvBKP"}]}]}} {"custom_id": "Jhenaidah_28215_14364_28728_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UdMHYccG49RPHr9aWSFo38"}]}]}} {"custom_id": "Bogra_20520_13338_21033_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-G9qqjkHCr3LFiDu7quBycm"}]}]}} {"custom_id": "Noakhali_7695_25137_8208_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Nr5FJY7hAFta1kcsUn4hhi"}]}]}} {"custom_id": "Bagerhat_14364_25650_14877_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VEBtExymDwbxPPUBMv4Bdg"}]}]}} {"custom_id": "Barisal_17955_18981_18468_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YLTxr8nBU4AXw2oVQ3Fa96"}]}]}} {"custom_id": "Maulvibazar_21033_10773_21546_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CN5CZN9nsqcQ6yZtnhxwM5"}]}]}} {"custom_id": "Sylhet_9234_23598_9747_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RY6X78ZYJN6y6r6ZrgMVBY"}]}]}} {"custom_id": "Brahamanbaria_24624_6156_25137_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7vvzEw9qdHYjviy4CPZnZW"}]}]}} {"custom_id": "Patuakhali_21033_16929_21546_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-G5NdYAdopVCBhdHiibGAqo"}]}]}} {"custom_id": "Pirojpur_9747_7695_10260_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NMk4T1pJmQKzdRpgn61Z2v"}]}]}} {"custom_id": "Bandarban_4617_20007_5130_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9qT3W1yXD5Kri6U3vdDEcR"}]}]}} {"custom_id": "Patuakhali_5643_13338_6156_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-12pnhmDQg5truc9pAYiT5E"}]}]}} {"custom_id": "Brahamanbaria_20007_15390_20520_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-89UU8FxKPDBo5FY3ARTi2U"}]}]}} {"custom_id": "Sylhet_8721_11799_9234_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3fUbEKKHbF5so88gWXKjic"}]}]}} {"custom_id": "Noakhali_12825_38475_13338_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9XhdZdrAX56Z3QxD15uLAY"}]}]}} {"custom_id": "Sylhet_9234_23085_9747_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9gQWmQaEwumEKPds28kC2x"}]}]}} {"custom_id": "Pabna_29754_24111_30267_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Pr9ddj3bM8nu7JeiYHGWRD"}]}]}} {"custom_id": "Pabna_15903_20520_16416_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JGc1s3JFxgh41YkPbUrjHh"}]}]}} {"custom_id": "Patuakhali_22572_6669_23085_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J1V95bE8zQJ2d8PR6Q1cVo"}]}]}} {"custom_id": "Chandpur_6669_15390_7182_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Jsx9o9ptJDEg9PhwhDngEv"}]}]}} {"custom_id": "Sunamganj_29241_20007_29754_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A4Ab2hJw9BvgHsUPRQNN8D"}]}]}} {"custom_id": "Sherpur_10773_16929_11286_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-64BfEc2SQ1RnJr6WHp1PzC"}]}]}} {"custom_id": "Kushtia_9747_5643_10260_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R6kXfafLXRggm7VjWRT84o"}]}]}} {"custom_id": "Rajshahi_14364_3078_14877_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QixjweQdN5MC2eoL4wWuvi"}]}]}} {"custom_id": "Chuadanga_14877_4104_15390_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-462YH4Z27YjJtmorzmrWGk"}]}]}} {"custom_id": "Khagrachhari_11799_8208_12312_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JRZA2TVaZ6e5NTZE8kXAus"}]}]}} {"custom_id": "Gazipur_12825_14364_13338_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1zR2Uhjt88ndkTAPUxyx76"}]}]}} {"custom_id": "Sunamganj_34371_17955_34884_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JXNi3kqFc62MudDC3aJexH"}]}]}} {"custom_id": "Naogaon_9234_1539_9747_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vd5H7ZLm9Fn1MTSw5TnLEY"}]}]}} {"custom_id": "Manikganj_10260_15903_10773_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AEowj66FJKc2cWS2VEnChh"}]}]}} {"custom_id": "Rangpur_19494_25650_20007_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NRuiY1Jm2BZCRriGuS1qcp"}]}]}} {"custom_id": "Bandarban_22059_17442_22572_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T6vx8XeDvv7qYLtppKWpTw"}]}]}} {"custom_id": "Sirajganj_20520_19494_21033_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DY7uAK3w8VT7m715gn7QrD"}]}]}} {"custom_id": "Khagrachhari_14364_22059_14877_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QoaxjQbJJ8C52Ln79EBaBg"}]}]}} {"custom_id": "Barguna_7182_19494_7695_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UuN1SoU5uS6dbZfHzvZyBk"}]}]}} {"custom_id": "Madaripur_8208_14877_8721_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JFeZyv6jqQv1ZPYLDD7a5F"}]}]}} {"custom_id": "Barguna_4617_19494_5130_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Xk6XSmCaD6JPfjhtx9Pw6o"}]}]}} {"custom_id": "Sherpur_7182_3078_7695_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X9iYKZzajT4DoPjBvbTt1Z"}]}]}} {"custom_id": "Satkhira_11286_42579_11799_43092.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1LwtZbojWJwKzsE7VVvqCw"}]}]}} {"custom_id": "Bogra_24624_26676_25137_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MDvNKKD1TSkDmnF8zDH7dt"}]}]}} {"custom_id": "Patuakhali_4104_12825_4617_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5q54yBJPXcpoZBDWmYwKPm"}]}]}} {"custom_id": "Jessore_24624_9234_25137_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U9pWNkKL1JTZ18VcQXv1WB"}]}]}} {"custom_id": "Thakurgaon_3591_21546_4104_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QNVsXT9EP5DoefHe3yaUt9"}]}]}} {"custom_id": "Chittagong_36423_41553_36936_42066.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6vUv6LB8cCkS7svitAbE78"}]}]}} {"custom_id": "Netrakona_25650_20520_26163_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R1oyX6wb4PqeRWV3GJXiWQ"}]}]}} {"custom_id": "Natore_7182_13851_7695_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4DyLFXAYgEnVGcaK2ohiYL"}]}]}} {"custom_id": "Satkhira_15903_7182_16416_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8mg8hXTLw87yVNhCMGAh8p"}]}]}} {"custom_id": "Maulvibazar_12312_20520_12825_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RDJLhfT4YL5CaXHoEvgxm1"}]}]}} {"custom_id": "Netrakona_29241_24624_29754_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TMk5RDyrQyBSZuQNCYEPfk"}]}]}} {"custom_id": "Patuakhali_3078_33345_3591_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-51EMs5KKt9cq7SdcGQcCCH"}]}]}} {"custom_id": "Narsingdi_11799_12825_12312_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J3Td6oN8MF1pt5LU8Ch5cC"}]}]}} {"custom_id": "Gopalganj_7182_22059_7695_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EaQEX1mmPfQbRNtkawZD53"}]}]}} {"custom_id": "Kurigram_13338_10773_13851_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-27Le873MvEjVwPCQYn2x63"}]}]}} {"custom_id": "Jamalpur_513_18981_1026_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fznmga9t9FDyfdechkTQnM"}]}]}} {"custom_id": "Gazipur_12312_13851_12825_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EPbUJyUxyWVC7iMjG1Eyz4"}]}]}} {"custom_id": "Kushtia_1539_513_2052_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BBfwgTWp7UbrVCRnuGQGJD"}]}]}} {"custom_id": "Madaripur_15903_2565_16416_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CNv985Komc2Sgiv3v7zcyV"}]}]}} {"custom_id": "Chandpur_6669_8721_7182_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7zLKGbTe3UUjUPBy1ZhsFQ"}]}]}} {"custom_id": "Rangpur_8208_18981_8721_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-51HRyVd7zmVqBxcV8aunPQ"}]}]}} {"custom_id": "Barguna_11799_9747_12312_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KfnpyYq2qRMTLxzePr5G8q"}]}]}} {"custom_id": "Chittagong_32319_38988_32832_39501.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5aFzBV24s1NKwhPgarG8eU"}]}]}} {"custom_id": "Gazipur_12825_513_13338_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1cqJHCzJkkDhZF5thtH8yZ"}]}]}} {"custom_id": "Kishoreganj_5130_14877_5643_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MfCz8VRGt3gcdtjGo9KM1J"}]}]}} {"custom_id": "Mymensingh_33345_30267_33858_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MiMy8RUaDtLUd8bmmgRRvr"}]}]}} {"custom_id": "Brahamanbaria_16929_23598_17442_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X2G8nEbxuV7ZCwnhQedzHV"}]}]}} {"custom_id": "Netrakona_14364_14364_14877_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LDzEkwJ25snKwZhXVciEjV"}]}]}} {"custom_id": "Netrakona_10260_11799_10773_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NqaoUDRizdCwhdeCe1VLGX"}]}]}} {"custom_id": "Comilla_21033_18981_21546_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2WD3zYLkpxerWo8jcimyyv"}]}]}} {"custom_id": "Khulna_8208_40014_8721_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BidcJvUYmgoLScdUTPBKDV"}]}]}} {"custom_id": "Khagrachhari_19494_28728_20007_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RVGreY9NkuCvmXRXUk49VN"}]}]}} {"custom_id": "Gazipur_10260_15390_10773_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XY5A7NQV1MTKZAGb8wiTHo"}]}]}} {"custom_id": "Pabna_30267_20520_30780_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-56TzytmhxnaioHpY9n9TuV"}]}]}} {"custom_id": "Netrakona_7695_11799_8208_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GntyXahDE6KuvAjXAMwZ4m"}]}]}} {"custom_id": "Mymensingh_13338_29241_13851_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CABej8Gn6xJYBng67ThYoq"}]}]}} {"custom_id": "Jessore_21033_26676_21546_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NYXSdXRk57ms7B4XhBX2ZF"}]}]}} {"custom_id": "Sunamganj_27189_7182_27702_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QTUZrwKMc6tkecX6BRwnzq"}]}]}} {"custom_id": "Comilla_7182_5130_7695_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ex2Q22M5Dy5Uu3UkESGF7G"}]}]}} {"custom_id": "Rajshahi_6669_4104_7182_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W8X9NZdPLZkTZsbh6M8pzG"}]}]}} {"custom_id": "Gopalganj_15903_6156_16416_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QsfP4n52yowbcHP823vUJ4"}]}]}} {"custom_id": "Maulvibazar_10773_14877_11286_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T8qGDs7goprfoHyjngK3s4"}]}]}} {"custom_id": "Munshiganj_9747_3591_10260_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Cv1zePQnbkT39QE4Y6jRFE"}]}]}} {"custom_id": "Mymensingh_6156_34371_6669_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CJ1D487aUnHX6Ep3P9a8JP"}]}]}} {"custom_id": "Comilla_25137_15390_25650_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4X666XV5owLTGsTkNt48Yj"}]}]}} {"custom_id": "Rangamati_15390_15390_15903_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8DD3zXrWonjMrieU6kCdWu"}]}]}} {"custom_id": "Barisal_18981_15903_19494_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QTtWnRAoNnpaXKyw7otKr8"}]}]}} {"custom_id": "Lalmonirhat_23598_18981_24111_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R4H3V5ZSmbN7q8aYUHxA7U"}]}]}} {"custom_id": "Lakshmipur_7182_16929_7695_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-17j8HqDWWq9dn2bt2qr1Qf"}]}]}} {"custom_id": "Chittagong_31293_31806_31806_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CSmd5Vp6sAWtXht5yQRVnT"}]}]}} {"custom_id": "Bogra_31806_15390_32319_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FNxkfBPNxycFViLzMfpUWT"}]}]}} {"custom_id": "Khagrachhari_11286_24111_11799_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vx5AVutbQd93jH6gsgk8S6"}]}]}} {"custom_id": "Dinajpur_14364_9234_14877_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FLtBzDiHVowvtqta68piE3"}]}]}} {"custom_id": "Lakshmipur_8721_11286_9234_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XJFbgSYYzytKkPhnGCsq5n"}]}]}} {"custom_id": "Magura_8721_1539_9234_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5QPXA5dEhK7Arw7eZW7b7t"}]}]}} {"custom_id": "Tangail_18468_13338_18981_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9HirDMjzW9r8jQg1Nhkuzd"}]}]}} {"custom_id": "Barisal_9234_2565_9747_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4pqNukGwPqMJxSMuhXkbKd"}]}]}} {"custom_id": "Khagrachhari_8208_33858_8721_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4td6Nw9sndcWX1LGE6VHRy"}]}]}} {"custom_id": "Gaibandha_24111_15903_24624_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9Bg4rNUSWfsbFPgKgyojsh"}]}]}} {"custom_id": "Brahamanbaria_17442_10260_17955_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-41vuSpokNKg2ef9kSKMVgj"}]}]}} {"custom_id": "Noakhali_513_4104_1026_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9vP2DjgvwUn6DSnPxmU43z"}]}]}} {"custom_id": "Kurigram_15903_15390_16416_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JwKHD7qXRLjziqFxKfvEUR"}]}]}} {"custom_id": "Chittagong_17955_4104_18468_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MHWbeJ3HpDgoXzJN4UA2Mi"}]}]}} {"custom_id": "Kishoreganj_11799_7695_12312_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-25gRjZDDdEPKamH9tUnEy8"}]}]}} {"custom_id": "Natore_18981_15903_19494_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4RJUa4fMbd2CfaJ3vCjkJd"}]}]}} {"custom_id": "Natore_2565_25650_3078_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-734CaeCb5g3tfgAfFpXZyL"}]}]}} {"custom_id": "Panchagarh_15390_18981_15903_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SYSGP3aXQzQiU5Xrgtjf9F"}]}]}} {"custom_id": "Bhola_17955_27189_18468_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A4wNXsymWLuq8XGXNzAsNt"}]}]}} {"custom_id": "Narsingdi_11799_3591_12312_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PRAFeokAuvZkbUJBxPvGZh"}]}]}} {"custom_id": "Sunamganj_31293_9234_31806_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vj1Q3WqyEY4zmH5hd4zoVY"}]}]}} {"custom_id": "Kurigram_7695_32832_8208_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HpYoKop4JHXYhqdyRboWXV"}]}]}} {"custom_id": "Gaibandha_14364_10260_14877_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4m8L2AVmcQm7yoHMjxNsFr"}]}]}} {"custom_id": "Khagrachhari_2052_23085_2565_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T62xRy5JHbGdZ412prSycF"}]}]}} {"custom_id": "Kurigram_13851_39501_14364_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2jxs61gk1PfRFjpj8XSp7y"}]}]}} {"custom_id": "Gaibandha_20007_29754_20520_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mh7rEAGUcVpLc3ByxhsK9b"}]}]}} {"custom_id": "Kishoreganj_4104_15390_4617_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ru53jj4on2jgBTP1trHvWP"}]}]}} {"custom_id": "Pirojpur_3591_5643_4104_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YJ6QCdXkRAs28JXqWhF8Fe"}]}]}} {"custom_id": "Rangpur_14877_14877_15390_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XPbhaBWHCY8oQjd9W71Evy"}]}]}} {"custom_id": "Noakhali_11799_24624_12312_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-52WLaeRGYBRofnCgiABfay"}]}]}} {"custom_id": "Kurigram_9234_25137_9747_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VkMn3fYL65DNXBoAw3f2D1"}]}]}} {"custom_id": "Chittagong_28215_22059_28728_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HSATbYec1pt8Tpf7U42kqp"}]}]}} {"custom_id": "Patuakhali_5643_38475_6156_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JTZoMW3jWULiRbvw8MrcJb"}]}]}} {"custom_id": "Noakhali_11286_10773_11799_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8L6yTzNo5xLNWzCvgXvqMk"}]}]}} {"custom_id": "Satkhira_10260_22059_10773_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AoP7JLPhjMTYYSCfZoXkUs"}]}]}} {"custom_id": "Gopalganj_6156_8721_6669_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6nN5iFy53GPkLun32Nnzgf"}]}]}} {"custom_id": "Gaibandha_14364_10773_14877_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GFg4EezaPEKhXkFDrzy3AG"}]}]}} {"custom_id": "Pabna_13851_7182_14364_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5ybgBV8yyN8KbA4bvXYAkm"}]}]}} {"custom_id": "Madaripur_12825_16929_13338_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CzfjxaUzcmby2rxUVyQbTS"}]}]}} {"custom_id": "Rangamati_14877_7695_15390_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hqq2czfBsNtvsPbtqRKCsw"}]}]}} {"custom_id": "Gazipur_16929_5643_17442_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Q5S5CSpZhh5VMFUrKSQ5tT"}]}]}} {"custom_id": "Thakurgaon_14877_25137_15390_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DFJDdG3t1bcN2MRiNgvQL1"}]}]}} {"custom_id": "Rangamati_13851_41040_14364_41553.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A3QKTR9DxpjfrgpBYMCorL"}]}]}} {"custom_id": "Bandarban_20520_13338_21033_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YU7Ear1to9mPust5rmN69U"}]}]}} {"custom_id": "Gaibandha_8721_27189_9234_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WhGUZhjvwtV1WaMCRf8Nt3"}]}]}} {"custom_id": "Joypurhat_8721_10773_9234_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CmpHqcv1UFTVtbF3AP9Tyw"}]}]}} {"custom_id": "Bagerhat_8208_38475_8721_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SUTDtKBxq63qxhWaaXFsqS"}]}]}} {"custom_id": "Maulvibazar_21033_10260_21546_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4ZaQA8JoGBofEAN5KdcZ6f"}]}]}} {"custom_id": "Kushtia_9747_6669_10260_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TtgHLgNa7pYXX9VfwsA65k"}]}]}} {"custom_id": "Brahamanbaria_15903_13338_16416_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A4sX4eT2TxRqjh57N2nQJ1"}]}]}} {"custom_id": "Jamalpur_2052_30780_2565_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FiQkYCcRDwbmfHxjDkG2Ub"}]}]}} {"custom_id": "Kurigram_13338_3591_13851_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8GQaxi9FWAvWgnVg41R2rv"}]}]}} {"custom_id": "Comilla_27702_30267_28215_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Nu3zK8n5C4dYatsaZHpHwK"}]}]}} {"custom_id": "Bhola_6156_4617_6669_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UAGLePAk2viPHqw6HgoyBj"}]}]}} {"custom_id": "Tangail_14877_29754_15390_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YTcEzdJK53i65HC7JrgtXR"}]}]}} {"custom_id": "Gaibandha_22572_11286_23085_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4pmmbtA1GBLpX7A7y3PN48"}]}]}} {"custom_id": "Chittagong_7182_10773_7695_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NerDSoTBWt2rwZzuRsXVCp"}]}]}} {"custom_id": "Noakhali_12825_4104_13338_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RQhApiESqGuxE4sKWtdKu5"}]}]}} {"custom_id": "Satkhira_4104_19494_4617_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7SE7Zpja41dESBRgCiv7Hj"}]}]}} {"custom_id": "Faridpur_5130_7695_5643_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SNgKiZUcUTaQfdoUCa1X3q"}]}]}} {"custom_id": "Satkhira_15903_51813_16416_52326.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3sUYuTd9HzqAzxDia6knQP"}]}]}} {"custom_id": "Kishoreganj_12312_9747_12825_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HrTxexMc15RfA3CK7K2Z8g"}]}]}} {"custom_id": "Naogaon_27702_9234_28215_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XYtxuHCZtHqRgpouhfmpJb"}]}]}} {"custom_id": "Tangail_6669_39501_7182_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7CFbda8jmZerLHWtiiaiRX"}]}]}} {"custom_id": "Barguna_6669_13851_7182_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6a5PuxW1T7ZvSSpTqcubZa"}]}]}} {"custom_id": "Kishoreganj_15390_14364_15903_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UKnQLKsM5YiRbBMFkFYSys"}]}]}} {"custom_id": "Bandarban_25137_43092_25650_43605.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TveUJJThDwqJKHuE8yzAef"}]}]}} {"custom_id": "Kushtia_29754_16929_30267_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KBwykcycmEV8UjYNJH4hns"}]}]}} {"custom_id": "Rangpur_17442_7182_17955_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EEdh3pMYruRwRDKrVrpaDS"}]}]}} {"custom_id": "Netrakona_5643_2565_6156_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-E1FGug85EgKuVBmBTdKGR4"}]}]}} {"custom_id": "Rangamati_14877_61560_15390_62073.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K8eDobuw8cvzy3n8a9zR1L"}]}]}} {"custom_id": "Munshiganj_1539_8208_2052_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6CwGB28m4EaYzVNGdvRHVg"}]}]}} {"custom_id": "Sunamganj_28215_16416_28728_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FmuELqVd8gnjvhFtsZJ2ai"}]}]}} {"custom_id": "Mymensingh_13338_1539_13851_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VXALo36AAVScm9UJjhMfcK"}]}]}} {"custom_id": "Mymensingh_15903_45657_16416_46170.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2rpVGgMA2vdhQHSCn7pvnL"}]}]}} {"custom_id": "Bogra_21033_7182_21546_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CcoCgCci6CZvdA7baRdLqo"}]}]}} {"custom_id": "Rajbari_15390_7695_15903_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FY65zap3NLoTDz9qgVhrCW"}]}]}} {"custom_id": "Mymensingh_27189_27189_27702_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8VrKH8MT58NpHp2K4QiU7a"}]}]}} {"custom_id": "Munshiganj_8208_3591_8721_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L3qPYPw2wDZywD4FUMekph"}]}]}} {"custom_id": "Maulvibazar_14364_27702_14877_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9E4g6dq9ztqCEZiAqTpJ8R"}]}]}} {"custom_id": "Mymensingh_14364_12312_14877_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HeAkM7z3x18D2AJ6qoVoS1"}]}]}} {"custom_id": "Naogaon_14364_3591_14877_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NnjcE3Af7z5BoDr3BDWDgn"}]}]}} {"custom_id": "Maulvibazar_19494_12312_20007_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FkZwCAz4tCN8dXxJypDPu5"}]}]}} {"custom_id": "Bagerhat_18468_26676_18981_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-21wNs3bSgS4JnfD4YWv5YZ"}]}]}} {"custom_id": "Sylhet_13851_24624_14364_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6aEtJCYwbmFuvEYr8ZKPXt"}]}]}} {"custom_id": "Gazipur_11286_9234_11799_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QDF4gW8EcBLPjW6mx93ksv"}]}]}} {"custom_id": "Bandarban_18981_41040_19494_41553.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M45khHSgv3qrCHJxzJEuHU"}]}]}} {"custom_id": "Chuadanga_12825_20007_13338_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YVeyUeCt4Gmf6Jhhurq1hA"}]}]}} {"custom_id": "Khagrachhari_13338_5130_13851_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Huozpuiyf5tcmdD9DcfF3G"}]}]}} {"custom_id": "Naogaon_28215_11799_28728_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KB31RWFszpUVB6Rd7X1M5d"}]}]}} {"custom_id": "Rangamati_18468_31293_18981_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Qc2GmS9TEnsjTitcV9rhHo"}]}]}} {"custom_id": "Bogra_28728_20007_29241_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WzS48R1ARuxQEXPw546oCa"}]}]}} {"custom_id": "Mymensingh_15390_34884_15903_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EsfQbtgjUW214wvQKX8Cww"}]}]}} {"custom_id": "Comilla_5643_8721_6156_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-923bRKceVFLGm7urvCLMMt"}]}]}} {"custom_id": "Rangpur_20007_16416_20520_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hw1bUGB3z7ovT168KS6cz8"}]}]}} {"custom_id": "Mymensingh_22059_36423_22572_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GSTY53GRU47dSKQJ42Ax14"}]}]}} {"custom_id": "Netrakona_21546_20007_22059_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MYbwToNDW31ENs4cWXaiit"}]}]}} {"custom_id": "Bogra_8721_12312_9234_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PB7fVqMyBdSH9QdkzNHS3i"}]}]}} {"custom_id": "Maulvibazar_6669_34371_7182_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FhNH8D6kTHzXq1boqY6FYX"}]}]}} {"custom_id": "Kushtia_15390_23598_15903_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7MmxnFJft8tT6L5cuFtJvx"}]}]}} {"custom_id": "Nawabganj_12312_9747_12825_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bmd9b8qZx1yxxcpr1AP3Jh"}]}]}} {"custom_id": "Comilla_23085_23085_23598_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TsX4JLeCszv8Uhxodqv8kq"}]}]}} {"custom_id": "Gazipur_15390_5643_15903_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EJhDUjVibDwUY1bx7gvRK8"}]}]}} {"custom_id": "Pabna_3078_11286_3591_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-65ZypZBNiJBHkjiJFZF5eQ"}]}]}} {"custom_id": "Tangail_13851_31806_14364_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BRgww77UGAAZvUzGa9zxET"}]}]}} {"custom_id": "Khulna_5130_58482_5643_58995.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vka8AfoNzTAHD1ji6sZqm7"}]}]}} {"custom_id": "Mymensingh_9747_46170_10260_46683.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PgFuSDJ4dHtAguLwPc9dfS"}]}]}} {"custom_id": "Barisal_12825_29754_13338_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4efW1sAr7nTS24WdfQLg6w"}]}]}} {"custom_id": "Rajshahi_15390_6156_15903_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W7RP5MGFzz3Uig3KtRA7t7"}]}]}} {"custom_id": "Chuadanga_4617_17955_5130_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KPTqobaorWMRqveKkYtLR4"}]}]}} {"custom_id": "CoxsBazar_12825_34884_13338_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R1KqCDzhNrFG1eU6mNBUGq"}]}]}} {"custom_id": "Rajbari_15390_9234_15903_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BkaL1PpVnT49L7xSF9YM5E"}]}]}} {"custom_id": "Narsingdi_13338_14877_13851_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GnzFVqrs12SJPAgEcSVMDr"}]}]}} {"custom_id": "Mymensingh_13338_42579_13851_43092.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3aiXugW9pZn1dtSdPxFymb"}]}]}} {"custom_id": "Jhenaidah_17955_6156_18468_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3b59tzhFyyqnPHbXVRNweg"}]}]}} {"custom_id": "Sunamganj_22572_10773_23085_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AeSk4BRfbt5ZdqCGmayJLx"}]}]}} {"custom_id": "Sirajganj_24111_7182_24624_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R9SvZhZZvKWAPDTUesPjht"}]}]}} {"custom_id": "Comilla_14877_12312_15390_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-26ccCrpC6fyuL1WzNegoNA"}]}]}} {"custom_id": "Barisal_18468_10260_18981_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BtQEv1z7fk577W77T4X3XZ"}]}]}} {"custom_id": "Satkhira_6156_27189_6669_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NdRX3pYAJzRCCbiUtWGZiF"}]}]}} {"custom_id": "Rangamati_15390_24111_15903_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KZkkcbqf9G2xCxZs89cTzU"}]}]}} {"custom_id": "Kushtia_15903_23598_16416_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X3vf3K7xo1XjPJMKBX1h9Z"}]}]}} {"custom_id": "Nilphamari_16416_9234_16929_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VZ4dzsWvaCDP8b9jsy4LRm"}]}]}} {"custom_id": "Tangail_3591_20520_4104_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3ThPwgzSZZWsaU28d9zyqt"}]}]}} {"custom_id": "Munshiganj_13338_7182_13851_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FuWLt1THhEyipm9h36F2DL"}]}]}} {"custom_id": "Rangpur_3591_16929_4104_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M3yonLJJVeMavLVsPaNS1z"}]}]}} {"custom_id": "Satkhira_11799_26676_12312_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UgLAtha9KRYJz2G6YzyGoY"}]}]}} {"custom_id": "Noakhali_11286_25137_11799_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-57WuV42S1u6rSUVHSgCsqT"}]}]}} {"custom_id": "Naogaon_31806_32319_32319_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FQmewJmg3giik2sbUKCFEB"}]}]}} {"custom_id": "Satkhira_7182_9747_7695_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RoJJ5dck9Sg6VBpNLagjKz"}]}]}} {"custom_id": "Nawabganj_12312_6156_12825_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D7jjSQtskoQYq3BaJkToK8"}]}]}} {"custom_id": "Bandarban_11799_7695_12312_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Kz2xbwKKLY6uUHAsAJhKNG"}]}]}} {"custom_id": "Panchagarh_21546_25650_22059_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ma2gHrKVfeNP7943rPvGec"}]}]}} {"custom_id": "Pabna_10260_18468_10773_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CP2S29m8KmwesZ4nXhsjfy"}]}]}} {"custom_id": "Chandpur_11799_17442_12312_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-349VZ3hUSAXo2RDY5jahBh"}]}]}} {"custom_id": "Satkhira_8208_41553_8721_42066.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-21jYYfpaeSFbu4WejLPh9L"}]}]}} {"custom_id": "Chittagong_25650_44118_26163_44631.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2L4uBaU61spHnkQBiiehEf"}]}]}} {"custom_id": "Gazipur_13338_0_13851_513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-H9MeTtBZgpo1XP1xFEmBVw"}]}]}} {"custom_id": "Madaripur_9747_19494_10260_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Py67V8sDwx8GUrmmZcex6z"}]}]}} {"custom_id": "Sylhet_10773_26676_11286_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UmvABo33EzVS2jUCnNaHfG"}]}]}} {"custom_id": "Tangail_2565_28728_3078_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Dh67YaywW8LSr7Erkmuodf"}]}]}} {"custom_id": "Comilla_25137_10260_25650_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WH13WLrmox8g5hjuG1g67P"}]}]}} {"custom_id": "Satkhira_20520_51300_21033_51813.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K2wBvFJPsrZGXHR22FUobJ"}]}]}} {"custom_id": "Rangpur_16416_30780_16929_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VzH68Nz7dopvvPNCMBDF2i"}]}]}} {"custom_id": "Rangamati_11286_63612_11799_64125.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Uv9HRVYj3xtdZnGJjBmxVc"}]}]}} {"custom_id": "Manikganj_16416_13338_16929_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7ic5uNiDqf7RSXv9q24nxb"}]}]}} {"custom_id": "Gazipur_10773_22059_11286_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VFfFeTw23YNzi7Nw4js4AN"}]}]}} {"custom_id": "Gaibandha_14364_5643_14877_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XEjXDMBzYoCnqPh8ye6wNg"}]}]}} {"custom_id": "Khagrachhari_3591_40527_4104_41040.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2xFckaQiXxP7j1zhGsBu6N"}]}]}} {"custom_id": "Lakshmipur_3591_6156_4104_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KL8st5n4JtuKJePhDcQ1nJ"}]}]}} {"custom_id": "Khagrachhari_6156_28215_6669_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WPTN7f6XF6hwocxQYLtywA"}]}]}} {"custom_id": "Maulvibazar_27189_16416_27702_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Qe7cCh3Hw9PLz48igMR83o"}]}]}} {"custom_id": "Netrakona_23085_14364_23598_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NvmgtZ4wedhKayRthMt8G1"}]}]}} {"custom_id": "Jamalpur_5643_6156_6156_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6y2GBeBB6UNM896s4kimgK"}]}]}} {"custom_id": "Maulvibazar_4617_17442_5130_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2aGhw9ZnXHLotjwgEmRU98"}]}]}} {"custom_id": "Narail_7182_8208_7695_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L8utagGwjrSohzDY4QKRNE"}]}]}} {"custom_id": "Mymensingh_25650_29754_26163_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QS1Gmq2BUMNhNfaJpatpaw"}]}]}} {"custom_id": "Panchagarh_15390_20520_15903_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GEj6579BmeaorJe1fkK9P6"}]}]}} {"custom_id": "Sirajganj_15903_13338_16416_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9gpXU9oDSTihvQuZTgWJcB"}]}]}} {"custom_id": "Rajbari_13851_5130_14364_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YJVvN6H7KvxH2x6jBmVo3W"}]}]}} {"custom_id": "Shariatpur_12312_6156_12825_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5p4CuSTh811qNLvXfYemSc"}]}]}} {"custom_id": "Rangamati_25650_70281_26163_70794.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KnZT5Y6wsCr1osSL8o4jGC"}]}]}} {"custom_id": "Chandpur_18981_9747_19494_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EhK3yVo62KxJRxkEwjwmBG"}]}]}} {"custom_id": "Dinajpur_15390_27189_15903_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-56FddqrUKgXy1RMd2tWHH1"}]}]}} {"custom_id": "Khagrachhari_12825_9747_13338_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MQkjSKRpj3ZMtTs2EYpJ87"}]}]}} {"custom_id": "Pirojpur_8721_10773_9234_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T1SGt98LU6xmdW6CRVKNSW"}]}]}} {"custom_id": "Shariatpur_13851_18981_14364_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-84KY57ffv23qKdxy5gi2kq"}]}]}} {"custom_id": "Habiganj_22059_18468_22572_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M8TgbjtNetWY8g4zzx5G13"}]}]}} {"custom_id": "Jessore_17955_1539_18468_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NEXxyMRpxawqV5sXHoVNt4"}]}]}} {"custom_id": "Bandarban_26676_51813_27189_52326.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9WqUp3YqEwPqqAe6MPKfn2"}]}]}} {"custom_id": "Shariatpur_12312_5130_12825_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VwEsqPdhaVWeWxLTzRLGdC"}]}]}} {"custom_id": "Jessore_23598_4617_24111_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9QAD7CrEWPpuin7hdtWP2p"}]}]}} {"custom_id": "Chuadanga_8208_22059_8721_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9QU1xLnsSkW3j3Rjib3bke"}]}]}} {"custom_id": "Bandarban_22059_21033_22572_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VqyMmawj9etHnyQT619qX3"}]}]}} {"custom_id": "Natore_13338_25650_13851_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7m81bjwSgMeS75KYMLUXZn"}]}]}} {"custom_id": "Munshiganj_10773_9234_11286_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CvVxcGBtSEExZ8KtmTLEni"}]}]}} {"custom_id": "Kishoreganj_17955_5643_18468_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-16k8mSRNyAhWDCAiRm1BLw"}]}]}} {"custom_id": "Mymensingh_24111_23598_24624_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Tzrkj9rvdm8XqXFBkcNKAZ"}]}]}} {"custom_id": "Satkhira_9747_29241_10260_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4YQDzzHMfyBs1e2BTwegjT"}]}]}} {"custom_id": "Chuadanga_12312_513_12825_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JYQWYoGLp4QCsZJHaYxc5s"}]}]}} {"custom_id": "Kurigram_16929_30780_17442_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NbnBaDAYhfNFcBidf9AzKC"}]}]}} {"custom_id": "Dinajpur_35910_35397_36423_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TdziVdidwJRJNMSvd93iYr"}]}]}} {"custom_id": "Satkhira_17442_53865_17955_54378.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ab7WMjihPBhZeZUqvaVjNi"}]}]}} {"custom_id": "Patuakhali_7695_37962_8208_38475.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QvNsarcJvgF1GhtR5hauJi"}]}]}} {"custom_id": "Gaibandha_13338_10260_13851_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5dbzBEZmcsZBMxiJ8iHsjW"}]}]}} {"custom_id": "Panchagarh_1539_18468_2052_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FUeqLEsRwKLbKhUu47r8xU"}]}]}} {"custom_id": "Sylhet_8208_4617_8721_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CgU3iVxrLuajf8WxB1FrmF"}]}]}} {"custom_id": "Chittagong_33858_31293_34371_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WrZqhkNxLV2oY7ztQ9yrE9"}]}]}} {"custom_id": "Khulna_24111_3078_24152_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DgSKbHSQ178wq8uJeQAH3j"}]}]}} {"custom_id": "Faridpur_18468_3078_18981_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1icZUjL9VvjsCkpv8qM2qH"}]}]}} {"custom_id": "Nawabganj_9747_23598_10260_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JjB2RxMPnNTUL3cxVLVEuf"}]}]}} {"custom_id": "Mymensingh_19494_40527_20007_41040.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VU2GLeiQm2oqspPqVUgzsb"}]}]}} {"custom_id": "Kushtia_5130_9234_5643_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J5y2MVqdoyBzWs6vfVsftd"}]}]}} {"custom_id": "Chandpur_6669_2565_7182_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V95EkKVhhP6YxwM1v5LvqW"}]}]}} {"custom_id": "Mymensingh_25650_27189_26163_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Y5JPBWfrEbFsZ8fcAnDm7e"}]}]}} {"custom_id": "Dinajpur_24111_25137_24624_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-25yR9pMXtPvppcC2fDBGzH"}]}]}} {"custom_id": "Kurigram_13851_43092_14364_43605.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4GXZMywUBySyRWpfRD1PiM"}]}]}} {"custom_id": "Maulvibazar_10773_30267_11286_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QEbLXyw6Lp3c7pN7nabPBm"}]}]}} {"custom_id": "Maulvibazar_1539_14364_2052_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-75EjdRZPK4XCFJ4RTCsfVe"}]}]}} {"custom_id": "Thakurgaon_5643_15390_6156_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Uzf4Bh14cEnqCLZKrNBprt"}]}]}} {"custom_id": "Chittagong_6669_21546_7182_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RmWvpSMTLfwY6NPEamAyER"}]}]}} {"custom_id": "Khagrachhari_7182_46170_7695_46683.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9RAPCa3CnCHkZ8Y3SCXGT4"}]}]}} {"custom_id": "Narail_13338_10260_13851_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8usTA6rF8QvRwT9ohbnsGb"}]}]}} {"custom_id": "Tangail_23598_18981_24111_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AVbFV4KpqVbHJQbXcgLqwZ"}]}]}} {"custom_id": "Sherpur_13851_11799_14364_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7KiZtDMh551ESFXwJ8vwvR"}]}]}} {"custom_id": "Barisal_1539_10260_2052_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-B2WM8UbcfKWegqQF1GSeAD"}]}]}} {"custom_id": "Sunamganj_6669_1539_7182_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RJ2ZVGFPdCLJNjAtq52LYa"}]}]}} {"custom_id": "Pabna_28215_15390_28728_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Kprm3h8awfq1WtCMSzFYLh"}]}]}} {"custom_id": "Jessore_17442_5643_17955_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A71LaHwhHj4jCJjSNNTwLt"}]}]}} {"custom_id": "Jessore_22572_14877_23085_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PozNqTumDqdvys4opAD5nT"}]}]}} {"custom_id": "Pirojpur_9747_8208_10260_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AWfyD1nN83hqamgSxg6QHc"}]}]}} {"custom_id": "Lalmonirhat_16416_18981_16929_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6k8FjD1Lvh3ShP6mph2ALL"}]}]}} {"custom_id": "Khulna_18468_4104_18981_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NX3z8SeCrEhf5Me5S8rC52"}]}]}} {"custom_id": "Tangail_21033_27189_21546_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RVvmbjwDbSbLsiXmnSEoGA"}]}]}} {"custom_id": "Rajshahi_22572_10260_23085_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BQBKVfHyNTihqckCCh7Swe"}]}]}} {"custom_id": "Dinajpur_10260_13851_10773_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3vECb82NmZeJQ3VkNZF3od"}]}]}} {"custom_id": "Jamalpur_10260_29754_10773_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NTTz9i5caTdR1e42AuEZCs"}]}]}} {"custom_id": "Khagrachhari_17955_8721_18468_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ALrGiHekSDveZHsTyugFRH"}]}]}} {"custom_id": "Comilla_3591_16929_4104_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-H3xV7iwoxhfWf4J3rnFqzn"}]}]}} {"custom_id": "Natore_10773_22059_11286_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D8ggbyHrenkDHLcnAVBaTu"}]}]}} {"custom_id": "Khagrachhari_16416_16929_16929_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QLk7t94BKrWsJeNRiMCbEk"}]}]}} {"custom_id": "Rajshahi_28728_3591_29241_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WKiXaRJCVuoUst2KS9oY6e"}]}]}} {"custom_id": "Rangamati_24624_58995_25137_59508.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LQjYjc6hxbG1ZPeFpkwnfB"}]}]}} {"custom_id": "Maulvibazar_14877_8721_15390_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-97gTPe42AZ5RrwLWHJiH8X"}]}]}} {"custom_id": "Mymensingh_9234_8208_9747_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PuwrNB6otzgp72Lkb8sHXd"}]}]}} {"custom_id": "Shariatpur_5643_15390_6156_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Nk32uS9mmMKYd4cBGFoNaA"}]}]}} {"custom_id": "Barisal_17955_26163_18468_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VfNhGnzmQmPmVzKrdPoykx"}]}]}} {"custom_id": "Sherpur_6669_20007_7182_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1vn5j586eT5zWoTSUA6K73"}]}]}} {"custom_id": "Mymensingh_19494_19494_20007_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A32oC3VvZexmzBC76PayaC"}]}]}} {"custom_id": "Jamalpur_9234_15903_9747_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F5UwYbx4zWk4yu7JKUJe2H"}]}]}} {"custom_id": "Jamalpur_20520_32319_21033_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JiG95Msem46gXokL41tT2c"}]}]}} {"custom_id": "Chandpur_12312_15903_12825_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PtFePGQ8Le6wTTHBAxnXzB"}]}]}} {"custom_id": "Bhola_15390_20520_15903_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PQYkuH4rt6kKkRwjvgBPU5"}]}]}} {"custom_id": "Pabna_11799_4104_12312_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YD5wZ4Xdq2E4MdJSpnLg76"}]}]}} {"custom_id": "Panchagarh_9234_19494_9747_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AXhEh87ASEzfyvUJQ61gBR"}]}]}} {"custom_id": "Chittagong_1539_22572_2052_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PRZr9oq9d4tQ3XJ48Ji3KJ"}]}]}} {"custom_id": "Habiganj_21033_10260_21546_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BTeh8rzPQNupomS1QEfXVu"}]}]}} {"custom_id": "Narayanganj_7695_10773_8208_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ShJJPr1hPX2yG7kjbtP8Fc"}]}]}} {"custom_id": "Sylhet_4104_5130_4617_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Rm1Qmtfi5ECFgYpBGH5yjG"}]}]}} {"custom_id": "Bagerhat_8721_14877_9234_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PTEyvNa1TbNyNiCnGRuHr4"}]}]}} {"custom_id": "Rajbari_20520_10773_21033_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bc9aa1SNSXcf2nGwZFgWeq"}]}]}} {"custom_id": "Mymensingh_8721_35910_9234_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M8qTz5CKnUTFTmGiqdeMgs"}]}]}} {"custom_id": "Kishoreganj_15903_22059_16416_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9qFtYpnAvUxXWfzwZeTnrm"}]}]}} {"custom_id": "Patuakhali_5643_39501_6156_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SSDAG52oiWsBgmzBwoyKz3"}]}]}} {"custom_id": "Sirajganj_12312_27702_12825_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UbkDtyE6yzpGyCaZpMddw1"}]}]}} {"custom_id": "Jhenaidah_22572_14364_23085_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BF1jETtiMdPKkyxeL93iW5"}]}]}} {"custom_id": "Jamalpur_5643_9747_6156_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TbSkYEv3Bg58MpJz4S79Uy"}]}]}} {"custom_id": "Satkhira_13851_34884_14364_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UeWHsNqx8qyid8oSNiwYfD"}]}]}} {"custom_id": "Lalmonirhat_27189_28728_27702_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BuyZHZn1LNa4oS5Smdjgjr"}]}]}} {"custom_id": "Natore_13338_0_13851_513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6TAcoFmpgCf4a9SwVBJ33A"}]}]}} {"custom_id": "Chittagong_32319_27702_32832_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3boDaG5MycPFA2vEf51bLM"}]}]}} {"custom_id": "Satkhira_8208_2565_8721_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R6kcyidYTpJb9KDgF9aADc"}]}]}} {"custom_id": "Sunamganj_27189_27189_27702_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UVvFTy3Yhf91ptmZvekB1B"}]}]}} {"custom_id": "Jessore_15903_12825_16416_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JinYj8nVhnuh6Wgx9dLqxw"}]}]}} {"custom_id": "Lakshmipur_3591_6669_4104_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Cjcsxd1XdgC4hKghFxmZpU"}]}]}} {"custom_id": "Bagerhat_7182_60534_7695_61047.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1LxA99d37ZmqboTZRoBjsP"}]}]}} {"custom_id": "Narayanganj_6156_18468_6669_18478.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HijFqVK5EptLCsUiF5L1vr"}]}]}} {"custom_id": "Maulvibazar_23085_20520_23598_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BrmmTCthMsFubF8GABNGjx"}]}]}} {"custom_id": "Kurigram_18468_35910_18981_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GbE5hz6qPVJ4jW5W4TUDZc"}]}]}} {"custom_id": "Satkhira_11286_28728_11799_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RnKrujTbKJEiJM6sN4kfMH"}]}]}} {"custom_id": "Magura_7695_3591_8208_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MgBwGFBPn8SpF32rKQGhCV"}]}]}} {"custom_id": "Bogra_17442_19494_17955_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LLPxRYdGukeQ5fmYAqcN4h"}]}]}} {"custom_id": "Comilla_18468_15390_18981_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KtSxuRSG1d5zV8KACtCfKb"}]}]}} {"custom_id": "Nawabganj_12312_11286_12825_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LuPhkgVVQjZ5299pc55nYL"}]}]}} {"custom_id": "Mymensingh_21546_46683_22059_47196.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UNYdkP9yb7AMmgdJNLMkGh"}]}]}} {"custom_id": "Nilphamari_9747_13851_10260_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KvWjy6EYbwy4eNT5pEwN8T"}]}]}} {"custom_id": "Feni_8721_18981_9234_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SsrJk8yvCzg8C7rr4MC2fn"}]}]}} {"custom_id": "Noakhali_9234_39501_9747_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S14gM6NJ9dSkauaH84eQAP"}]}]}} {"custom_id": "Khulna_6156_49248_6669_49761.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TqqaSeaFEi6TyJfNvQhokK"}]}]}} {"custom_id": "Bandarban_14877_40527_15390_41040.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9rk9x29P6MTagnZq64iiFC"}]}]}} {"custom_id": "Jessore_22059_26163_22572_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8GdoM3ewSRW6CPVjvJ889i"}]}]}} {"custom_id": "Dinajpur_11286_4104_11799_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XAC6QbHdpzSXoSGYw7S2Lj"}]}]}} {"custom_id": "Rajshahi_25650_30267_26163_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U6C7hqEV6wGbMmPuxVzpn3"}]}]}} {"custom_id": "Khagrachhari_11286_34371_11799_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VYv1FCgJ9FFk93k2nPADcP"}]}]}} {"custom_id": "Bogra_20520_15903_21033_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LYygWKwgzBh8R2Z64d4kgY"}]}]}} {"custom_id": "Shariatpur_513_7695_1026_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PCDBjWpp3jCeaNoY3Crok8"}]}]}} {"custom_id": "Gazipur_7182_2052_7695_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9WMv2jmnJXZV2G1u41LGx9"}]}]}} {"custom_id": "Khagrachhari_3078_35910_3591_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TvB3vkULYphD9JEPcjx83t"}]}]}} {"custom_id": "Bogra_19494_26163_20007_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EUCCNcq42ytoBs68ae3AhM"}]}]}} {"custom_id": "Joypurhat_6156_10260_6669_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WKtCwErPGhTZMfoGDaRz1m"}]}]}} {"custom_id": "Barguna_11286_17955_11799_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2QKX1WfiW49NJVTruA9w6r"}]}]}} {"custom_id": "Pirojpur_9234_4104_9747_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8rEWg1XKH5twuvyMpEGnpH"}]}]}} {"custom_id": "Bogra_12825_6156_13338_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-48JRwYfCAZ1skXDzLhDHzz"}]}]}} {"custom_id": "Panchagarh_6156_20520_6669_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QWT6eSdCfVuB9sAW6yJxhP"}]}]}} {"custom_id": "Pabna_27702_21546_28215_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8rKZFCKNCHBEY91EGPLkNP"}]}]}} {"custom_id": "Brahamanbaria_24624_2565_25137_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7hxQQSaNkW2C53aAjaJ1fn"}]}]}} {"custom_id": "Bogra_30267_9234_30780_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Dj6rxFY6T2sGutvKYdLFQd"}]}]}} {"custom_id": "Comilla_24111_20520_24624_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NUawP9PQu9XshZxzFv7rNo"}]}]}} {"custom_id": "Bhola_9747_33858_10260_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LyQXHKZS3kh4KhJBu5Lxgq"}]}]}} {"custom_id": "Gazipur_17955_13338_18468_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Txm6z7CQJSSmJ1JWXhdnZi"}]}]}} {"custom_id": "Chandpur_11799_20520_12312_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HaQhYHcsoekKRSmiqxceYy"}]}]}} {"custom_id": "Faridpur_18468_14877_18981_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VnMqNprtYko3EKprgKoXcY"}]}]}} {"custom_id": "Sirajganj_11799_23598_12312_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7p8xR5b8pgibNMBzYNWY1X"}]}]}} {"custom_id": "Tangail_16929_18981_17442_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LsDC2F7W9fLVVVXtsWPgcY"}]}]}} {"custom_id": "Khagrachhari_11799_33345_12312_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R36SigPVdEz2B5bgzNyr9y"}]}]}} {"custom_id": "Kurigram_8721_7182_9234_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CeGVbh1sqKxek88LdxRxp3"}]}]}} {"custom_id": "Gopalganj_5643_10773_6156_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VAavFA838MXqSguDxsQ89p"}]}]}} {"custom_id": "Faridpur_19494_3591_20007_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5RkrS6GwuMhwkM8geft42o"}]}]}} {"custom_id": "Patuakhali_17442_5130_17955_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L6CuLP2nYh3b8ueffC9PJu"}]}]}} {"custom_id": "Rajshahi_25650_26676_26163_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6FCeern4HN3uRu3RtwcaGs"}]}]}} {"custom_id": "Narsingdi_11286_4617_11799_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XZATsiBwoccfFrZCsGy8gs"}]}]}} {"custom_id": "Satkhira_1539_1539_2052_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ArSus4mQUMTsiHEW2fJVCg"}]}]}} {"custom_id": "Barisal_21033_22572_21546_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5X3CZpBRRFTwonBSTFVHMW"}]}]}} {"custom_id": "Mymensingh_2565_21033_3078_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HC4fCYaTqdmY4AYxoLLPdy"}]}]}} {"custom_id": "Lakshmipur_14877_21546_15390_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WnTZtSc2zByvuaAaVN2WXk"}]}]}} {"custom_id": "Noakhali_2565_1539_3078_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VfxNgnTe7Dh9uKPNmvX3sk"}]}]}} {"custom_id": "Tangail_7182_36423_7695_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BuwdNNd8PcwJTv1H6LmfN7"}]}]}} {"custom_id": "Kishoreganj_24624_12312_25137_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GvJnzXRY3UBpS4W2JCvF1T"}]}]}} {"custom_id": "Gopalganj_7695_12825_8208_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6M6CHAeCbzGH8iBjWcm1L5"}]}]}} {"custom_id": "Satkhira_6669_29241_7182_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X3PMfqLaxEJ7TqtTAwG4dg"}]}]}} {"custom_id": "Satkhira_17442_37449_17955_37962.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MaY6WL4C4nSnu46ks1LiC5"}]}]}} {"custom_id": "Kishoreganj_21546_27189_22059_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VhHe3yzL5JJUzj2Mp9zBiK"}]}]}} {"custom_id": "Naogaon_12312_26676_12825_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M6U75rYZZLtdbF3Rt91kah"}]}]}} {"custom_id": "Barguna_13338_24111_13851_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CWtU3KHGGaumeVCwg5QrDx"}]}]}} {"custom_id": "Netrakona_26163_29754_26676_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MbpDYojyGdD4pYwUoNuYqe"}]}]}} {"custom_id": "Kishoreganj_30267_10773_30780_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1j2nkz3s5FFhYYtsMnFxHH"}]}]}} {"custom_id": "Nawabganj_13851_19494_14364_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WBcpVuPzqeoeR7kUSxwzJ2"}]}]}} {"custom_id": "Rangpur_2565_10773_3078_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EeFMajKHoPCQw3DWkgcBqs"}]}]}} {"custom_id": "Brahamanbaria_9747_23085_10260_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RUGMMkjiiqDMjXikH1W8ko"}]}]}} {"custom_id": "Maulvibazar_13851_32319_14364_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2kWA3D7DWTGqrktxSPywap"}]}]}} {"custom_id": "Sylhet_28215_6156_28728_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ya5FjUxocdfQFyEGXm71jz"}]}]}} {"custom_id": "Kushtia_14364_12312_14877_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5jypMGT9raevzbLb7D8UsB"}]}]}} {"custom_id": "Jamalpur_15390_29754_15903_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JeECXAuC2BFgvRD9tGJTUf"}]}]}} {"custom_id": "Sirajganj_10773_15903_11286_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WXURodKghRm6tdbiARqe2V"}]}]}} {"custom_id": "Gaibandha_9747_25650_10260_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EtGfxNgk64FZYbSPtCFedd"}]}]}} {"custom_id": "Sunamganj_29241_16929_29754_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Wf8v6x5paq4y9qxtVenve6"}]}]}} {"custom_id": "Chandpur_1539_3591_2052_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LiCVxHM6HQaLyvPMPe98nJ"}]}]}} {"custom_id": "Satkhira_13338_36423_13851_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5Ft4ytLjxD1neSnQFxyxu8"}]}]}} {"custom_id": "Sherpur_2565_12825_3078_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ATU3atSUVfyAM9rncW6z93"}]}]}} {"custom_id": "Maulvibazar_14877_28215_15390_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HEsDGQwdpkFP5Mxe7wTYV6"}]}]}} {"custom_id": "Khagrachhari_11286_29241_11799_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LhJmEdXeZuCkaHKT9jgtsC"}]}]}} {"custom_id": "Jhenaidah_15390_7182_15903_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CF8HT6ocQXF1ukPi4yfHgy"}]}]}} {"custom_id": "Maulvibazar_8721_12825_9234_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HgySDdSRdjXnfaYQTaGVR8"}]}]}} {"custom_id": "Rajshahi_0_14877_513_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8yjGtF2qE9yEnxrMmAoPq4"}]}]}} {"custom_id": "Sherpur_16929_17955_17442_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CCM6URDnrJhKWrE5NoTV5R"}]}]}} {"custom_id": "Jamalpur_18981_26676_19494_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5qGvYUGkUfrgVpMQ9PvPiL"}]}]}} {"custom_id": "Kishoreganj_7695_17955_8208_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7hvKvqfchCjP5u1ggxMzw3"}]}]}} {"custom_id": "Maulvibazar_25650_13338_26163_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L21L3AcYNKTKAezvFcCjVY"}]}]}} {"custom_id": "Kurigram_17955_12825_18468_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RcYMcJtcf3dSgZRC8oM7rV"}]}]}} {"custom_id": "Naogaon_30267_33345_30780_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NTiz4gCroru2LD2iqSKpkG"}]}]}} {"custom_id": "Tangail_17955_29754_18468_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mtyr1x5HaXFFThXEqJAFa9"}]}]}} {"custom_id": "Dinajpur_10260_14364_10773_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VV36bc5TQ9ZYnZKmM7vUi9"}]}]}} {"custom_id": "Bogra_12825_2565_13338_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6Y1wSXqHCPJrDKkUmZ5QZD"}]}]}} {"custom_id": "Natore_6156_8721_6669_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3EcLe6mMYub4qfg3t18nXY"}]}]}} {"custom_id": "Jamalpur_3591_14877_4104_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FaxsaGgBzenq6hr5WdxG1C"}]}]}} {"custom_id": "CoxsBazar_9234_22572_9747_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LL7wDgjrM3D1axMRemrfTQ"}]}]}} {"custom_id": "Rajshahi_11286_8721_11799_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MnYdhsbNxrz1YXdVC8vK3H"}]}]}} {"custom_id": "Jessore_7695_17955_8208_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M9aLDRVvwyTigGvofyqXTn"}]}]}} {"custom_id": "Rajbari_2052_9234_2565_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YXDf5My6v5S5U6h8e5DN7J"}]}]}} {"custom_id": "Netrakona_24111_16416_24624_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NknYGJVR1LYmxc47T5Xb4v"}]}]}} {"custom_id": "Rajshahi_22572_3078_23085_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HRXhQHaALBjKMf96ZtsWmG"}]}]}} {"custom_id": "Rajshahi_26676_25137_27189_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VpaCVeVwWBi4bev9tZamM3"}]}]}} {"custom_id": "Khagrachhari_10773_40527_11286_41040.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WYYJEpaEdsSqkowq46VTYH"}]}]}} {"custom_id": "Satkhira_7695_9234_8208_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FCZr3JAvhQHPmwKxfsie5A"}]}]}} {"custom_id": "Netrakona_14877_12825_15390_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Lj2KFp7pgZttJpAmw1KrNi"}]}]}} {"custom_id": "Brahamanbaria_23598_25650_24111_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8eKHkt4iuQ3EzAbAAuyd4g"}]}]}} {"custom_id": "Mymensingh_7182_9747_7695_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Wq3yuiyiskbdvx1xXGc19N"}]}]}} {"custom_id": "Nawabganj_4104_16929_4617_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EAqRxT5QcxGYgvZRdav2KA"}]}]}} {"custom_id": "Sylhet_20007_15390_20520_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2K6PPU67iU4fXRC7wyfsfc"}]}]}} {"custom_id": "Sylhet_4617_20520_5130_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MRxCyhaRQsDXATY4G8bxdJ"}]}]}} {"custom_id": "Khulna_11286_27702_11799_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UWCZATswZ37dtRuRUyNZAK"}]}]}} {"custom_id": "Chuadanga_7695_15903_8208_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SBtpbMCTahaJpiH8n4QCu6"}]}]}} {"custom_id": "Chandpur_2052_23085_2565_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Va9JMyQby3fugjUPECYDDc"}]}]}} {"custom_id": "Khulna_6669_4617_7182_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FVaJxxPL8CsAiocDgFgzJx"}]}]}} {"custom_id": "Sunamganj_7695_1026_8208_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X7tu6hgN5JckQbgPniA8RR"}]}]}} {"custom_id": "Mymensingh_11799_14877_12312_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XrzcD7F4kGSmgAJocyNgbD"}]}]}} {"custom_id": "Tangail_22572_20007_23085_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UnTWAAMMpC5BnvGkCyFRBA"}]}]}} {"custom_id": "Kushtia_10773_16416_11286_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MUKQy9Ko989S6xcV9L761T"}]}]}} {"custom_id": "Faridpur_21546_7695_22059_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VsW7sUV8ornP5Vaj1Z6Q7q"}]}]}} {"custom_id": "Comilla_23085_34371_23598_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CMxHP8TJkkRB323bN6Vbqn"}]}]}} {"custom_id": "Naogaon_28728_32832_29241_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ry3vfFWeMC9G5vy1ak7Gwv"}]}]}} {"custom_id": "Rangpur_5643_10773_6156_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Nb54WNAf8pV4yW6meUH5nn"}]}]}} {"custom_id": "Bagerhat_15903_41040_16416_41553.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-C3XXfG2eLYUmE5gr8JuCKY"}]}]}} {"custom_id": "Comilla_22059_17955_22572_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LqqbKYsiPte8hSRR2m3xAj"}]}]}} {"custom_id": "Kishoreganj_24624_10773_25137_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S5hFQUPgoY4YEooCRZ9u5p"}]}]}} {"custom_id": "Thakurgaon_12825_15390_13338_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8K87hb7KtZ7DQnggcm9G4D"}]}]}} {"custom_id": "Gaibandha_16929_21033_17442_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EtzNmcgr5NWsDuCdHspRsu"}]}]}} {"custom_id": "CoxsBazar_17442_41553_17955_42066.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FX5DTT69x7C3gU92SBsuTZ"}]}]}} {"custom_id": "Nilphamari_15390_17955_15903_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-E6a622ZqVf4EgYcyHZoHpT"}]}]}} {"custom_id": "Kurigram_10773_25137_11286_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FwZBXw62EGsJZpY4QUaJqr"}]}]}} {"custom_id": "Bandarban_25137_42066_25650_42579.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vr1u9LjC4qsgZB6qZCPnEg"}]}]}} {"custom_id": "Faridpur_22059_10260_22572_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5hQffjVbyHjHp2tDCvMmBT"}]}]}} {"custom_id": "Panchagarh_3078_18981_3591_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N9qpNrAezYfmEAiRq8FVYV"}]}]}} {"custom_id": "Sylhet_14877_1539_15390_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MiXxASS8b1S4BsLSi5gQo2"}]}]}} {"custom_id": "Kushtia_21546_17442_22059_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JJi1zYsLm68A2jRuFLxWdk"}]}]}} {"custom_id": "Meherpur_4617_5130_5130_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FTCXRqaQWJGUVYv8erJX93"}]}]}} {"custom_id": "Brahamanbaria_1539_28728_2052_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GPsgGNrPDMuvwWvKst4C2x"}]}]}} {"custom_id": "Kushtia_19494_24111_20007_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2Bbn7DG5LAFQNaB3YzkZPx"}]}]}} {"custom_id": "Khulna_7695_3078_8208_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-49ZDnQMGQPxecLdd3jWT2C"}]}]}} {"custom_id": "Chandpur_18468_7695_18981_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3QdpkrDoskB6sMFuQFuv4P"}]}]}} {"custom_id": "Brahamanbaria_6669_25137_7182_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SW6eczGNiNBVsAKcMSci7V"}]}]}} {"custom_id": "Manikganj_15390_8208_15903_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XTgSxeq9RYQQbY2zv47oU4"}]}]}} {"custom_id": "CoxsBazar_13338_35910_13851_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8Cf8EYqNUqME15cJmRwAiY"}]}]}} {"custom_id": "Rangpur_18981_15390_19494_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M29h48jsK2a9yBtEkhNqQa"}]}]}} {"custom_id": "Magura_11286_7695_11799_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-No3MswAxtT7AQtEXuUDceJ"}]}]}} {"custom_id": "Sunamganj_8208_12312_8721_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UJpGiBm9znTe22VZ1b5HLx"}]}]}} {"custom_id": "Pirojpur_9747_17955_10260_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DDfipR3RDWbFWCi5sSfrHw"}]}]}} {"custom_id": "Bagerhat_2052_48222_2565_48735.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3qPqm2HNDmvmfhHMGemD7E"}]}]}} {"custom_id": "Netrakona_18468_30780_18981_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HVC5THkmPkLTYGJFsYDUUK"}]}]}} {"custom_id": "Lalmonirhat_2052_5130_2565_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K4dUVRxJAxy9DVRF65JWwr"}]}]}} {"custom_id": "Habiganj_22572_6156_23085_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7ca38ijuLLusPttKqWCaxN"}]}]}} {"custom_id": "Pabna_27702_21033_28215_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7tDFFatjNyVWFntK3JNW5w"}]}]}} {"custom_id": "Thakurgaon_11286_22059_11799_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CwYzVkQZL7pEuMAKbTURup"}]}]}} {"custom_id": "Pabna_24624_15903_25137_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-H4uCjWakaguz3znUQ3pGQT"}]}]}} {"custom_id": "Naogaon_34371_27189_34884_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-63aScs8P4Wop83DdK4YeJx"}]}]}} {"custom_id": "Chuadanga_10260_18468_10773_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-71kqVADD6EMDZvRSXg2AZP"}]}]}} {"custom_id": "Mymensingh_9234_21033_9747_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1ww8164mNFD5CBFDvN4T8k"}]}]}} {"custom_id": "Gazipur_19494_13338_20007_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7XkgERjdTs5YeShHzbKyGq"}]}]}} {"custom_id": "Rajshahi_16416_15903_16929_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CLV36up1U15HmTAWVWRBVn"}]}]}} {"custom_id": "Habiganj_12825_5130_13338_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LoMczmHqQrjH3Ev95frEBe"}]}]}} {"custom_id": "Bandarban_2052_28215_2565_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RYMzj1kBuYKwTXXaYEWAZq"}]}]}} {"custom_id": "Netrakona_14364_16929_14877_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PLC6ePswQUsnk5NjdWUL1e"}]}]}} {"custom_id": "Patuakhali_26676_10773_27041_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VF6RWbXr62tv5Vp4yii73Q"}]}]}} {"custom_id": "Sylhet_32319_11799_32832_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AVdwQubpdoP8Vjjmo55Abq"}]}]}} {"custom_id": "Mymensingh_9234_39501_9747_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9CvPonYoS2t1HuBv4vvrCU"}]}]}} {"custom_id": "Dinajpur_27702_24111_28215_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DRsYP8g6ib5tmsqGvn3jXW"}]}]}} {"custom_id": "Barguna_15390_17955_15903_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AztDPN9XPCxDbp25vEhuGD"}]}]}} {"custom_id": "Mymensingh_23085_21546_23598_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7AB14B9qjA12BQqZk4GuNu"}]}]}} {"custom_id": "Bhola_16416_24111_16929_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QgFERPv75hYrukA3N2xz7j"}]}]}} {"custom_id": "Bandarban_11799_44631_12312_45144.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-54rEtcwMjMrEurvM5XmUri"}]}]}} {"custom_id": "Mymensingh_20007_45144_20520_45657.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RmrLxeSJ5uP2dYzZEMXsgV"}]}]}} {"custom_id": "Khulna_12825_28215_13338_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fa5N92uBTVU1E7ANjiKxT2"}]}]}} {"custom_id": "Sunamganj_11286_7182_11799_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Jo2Soj24JbpJBEeBYgj1A9"}]}]}} {"custom_id": "Lalmonirhat_28215_31293_28728_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Xs5Et7euWgdfmBfELLcsXX"}]}]}} {"custom_id": "Mymensingh_1539_21033_2052_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WE654skwj4iePKjCBBF5bN"}]}]}} {"custom_id": "Gopalganj_10773_4617_11286_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EgVuMCdCffAEuaeeT2pkHs"}]}]}} {"custom_id": "Kishoreganj_17955_18981_18468_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TC7DRh6dyMx6ooM7Jcive2"}]}]}} {"custom_id": "Gaibandha_18468_7695_18981_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7Xkp6Ph1kDw4sADig9bZh7"}]}]}} {"custom_id": "Noakhali_5643_10773_6156_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-G1xbDe4MerTkfzxwxypsR5"}]}]}} {"custom_id": "Magura_11799_12312_12312_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5HB7LQq6CeKNybFomBTzH2"}]}]}} {"custom_id": "Bhola_11799_16416_12312_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6yyQ3Cysx5eBaWbV1qke24"}]}]}} {"custom_id": "Munshiganj_18981_7695_19494_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6PHq1MLWGQKmZDTqVYfAQW"}]}]}} {"custom_id": "Bhola_5130_13338_5643_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KfsXPoCthTsaisaXssMr8M"}]}]}} {"custom_id": "Thakurgaon_7182_18981_7695_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GzSAM5JFds4Ap1TSw1PRPE"}]}]}} {"custom_id": "Khulna_10773_20007_11286_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J8kBkwBUW2qjT9pcHUUQ3r"}]}]}} {"custom_id": "Barisal_13851_15903_14364_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-81GBUQRnNBxcehpkAaKLR1"}]}]}} {"custom_id": "Khulna_3591_29241_4104_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W8LMN2ms5nMKeVkz81qx6Y"}]}]}} {"custom_id": "Narayanganj_13338_8208_13851_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RVQp5Rx4rfV2WVxEC4bEBk"}]}]}} {"custom_id": "Brahamanbaria_22059_7695_22572_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KR1FTaLjqXbPv4yFUUpJDy"}]}]}} {"custom_id": "Narail_6669_9747_7182_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SRUsJD4YJ2EGdmhdP4sy5f"}]}]}} {"custom_id": "Sylhet_20007_18981_20520_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J7WZr9ZJKXYLRVLdRxffEL"}]}]}} {"custom_id": "Meherpur_8208_3078_8721_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AgMAgpirUspGYjy9ytmBVM"}]}]}} {"custom_id": "Nawabganj_14877_17955_15390_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VpskwtbvT6XijR7fzCRtCk"}]}]}} {"custom_id": "Jamalpur_7695_33345_8208_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-57G3eTKB1FsW5kCmBzzhHw"}]}]}} {"custom_id": "Narail_10260_12825_10773_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AKz1S5ExCZna5P7VMkR13u"}]}]}} {"custom_id": "Bagerhat_12825_16929_13338_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ABdZfm45NjPKwqMvvydP3w"}]}]}} {"custom_id": "Pirojpur_9747_19494_10260_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WLoKLp9zDZvfipxPbbNzyS"}]}]}} {"custom_id": "Chuadanga_8721_3591_9234_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3irBL4CoUxoR6GZNwyFXnp"}]}]}} {"custom_id": "Dinajpur_35397_32832_35910_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DtEh8WaaSt1GdjWu9dKJDP"}]}]}} {"custom_id": "Jamalpur_6156_36936_6669_37449.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VfSvfF6jYqhb4w2Bz4kNgo"}]}]}} {"custom_id": "Khulna_2565_26163_3078_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8ssjPKWRvYumEQDq92owwZ"}]}]}} {"custom_id": "Rangamati_15903_45144_16416_45657.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HmjeAKW6KuSLHPX9JvABFw"}]}]}} {"custom_id": "Rajbari_14364_7695_14877_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3SeGUfSKQVpdu1JJv5vmiN"}]}]}} {"custom_id": "Sirajganj_15390_13851_15903_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XTy5Q56thjXg5dvw7MveUj"}]}]}} {"custom_id": "Bogra_20007_29241_20520_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DAvdFZVNYcfKKnGah5RuHK"}]}]}} {"custom_id": "Thakurgaon_7695_14364_8208_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BBCAYnmQw3mdYMki7PnToj"}]}]}} {"custom_id": "Panchagarh_18981_15390_19494_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BJwVFyWN1pCyTCpj1andae"}]}]}} {"custom_id": "Lakshmipur_15390_10773_15903_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-War37ioP7CrhxoBGRQ9u6v"}]}]}} {"custom_id": "Chittagong_20007_21546_20520_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J3p413HQXKB9ijWv8dHwBx"}]}]}} {"custom_id": "Natore_15390_17442_15903_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7u4FJgYGNZayXSWo8xdAcV"}]}]}} {"custom_id": "Noakhali_17955_19494_18468_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DgYHtdkPLD71hSkuCqxeyv"}]}]}} {"custom_id": "Mymensingh_25137_29241_25650_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-G7XBTdmaiTsHYTFaJdsAtp"}]}]}} {"custom_id": "Kishoreganj_22059_20007_22572_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LN688iyq5R3U3GYb4kUwrB"}]}]}} {"custom_id": "Sunamganj_20007_2565_20520_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J4y7JxytK3WVSwnogYghCw"}]}]}} {"custom_id": "Rajbari_1026_9234_1539_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5s4rT55cQEHrPYurGxNU5i"}]}]}} {"custom_id": "Noakhali_1539_2052_2052_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EijifXDETi8Q6bw4zQ5PPn"}]}]}} {"custom_id": "Patuakhali_21546_24111_22059_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RJn82YRaSSRZMDq7PRzFMB"}]}]}} {"custom_id": "Gaibandha_15390_15903_15903_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8GkQaf33AUpEEkd3iSwpBa"}]}]}} {"custom_id": "Rangamati_21546_24624_22059_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YPJFwEgdFYP6tTKtQdSfuq"}]}]}} {"custom_id": "Kishoreganj_13851_5643_14364_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-52aNkhQcyrL1C1ZEW9AFgd"}]}]}} {"custom_id": "Noakhali_2052_3078_2565_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WnooSVzMvRkxt5896d9Zq5"}]}]}} {"custom_id": "Mymensingh_9234_32832_9747_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1DEnWx6TuHyQaxfwgb2Nt3"}]}]}} {"custom_id": "Jhenaidah_30267_6156_30780_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UNAoSLxyM2S18Ux8qRnwg5"}]}]}} {"custom_id": "Jhalokati_6156_4104_6669_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9E2GWtbnWjpHg8TUN7QaQx"}]}]}} {"custom_id": "Habiganj_18468_4104_18981_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LWAVmphyNuCYrmWA1q3U7T"}]}]}} {"custom_id": "Sirajganj_9747_16929_10260_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6bszYF2qzoePRxDE3TXNZW"}]}]}} {"custom_id": "Khulna_9747_36423_10260_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gb6P4asPccv1ddMSp48kZD"}]}]}} {"custom_id": "Bandarban_23598_23085_24111_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4v7yWbaWroAhUnDuqfk5Vm"}]}]}} {"custom_id": "Jhenaidah_16416_0_16929_513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XDsPqwGrMVj7qoDPruAqxY"}]}]}} {"custom_id": "Jessore_15390_21546_15903_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AqYWnF2cgdwwz9N33FuZ9K"}]}]}} {"custom_id": "Kushtia_7695_12825_8208_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TQexN8mRE9q7LR658XXmjs"}]}]}} {"custom_id": "Bogra_10260_20007_10773_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7NiQe2WD73HpJiTx5MyLok"}]}]}} {"custom_id": "Naogaon_20520_28728_21033_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GpGyDVAMi7TS3fQCFkFLzR"}]}]}} {"custom_id": "Pabna_15903_513_16416_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M8NjTqWPxfnwEkx5fzejfU"}]}]}} {"custom_id": "Jhalokati_8721_1539_9234_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5j2z7KnsQTGjYDu6uLSFqh"}]}]}} {"custom_id": "Rangamati_12312_40014_12825_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A5C6BFtBZSgg869hLeRSH3"}]}]}} {"custom_id": "Rangamati_7695_6669_8208_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RYD95ayqjo2zrfkKWTdvmA"}]}]}} {"custom_id": "Comilla_8721_15390_9234_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9sGkEZPHxirndBcGzQiY1V"}]}]}} {"custom_id": "Khulna_2052_34884_2565_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-547eFZhL7TqLBpHpFX8y4E"}]}]}} {"custom_id": "Gazipur_19494_13851_20007_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WpBMntJZFXmtQckoTcec3R"}]}]}} {"custom_id": "Meherpur_11286_3078_11799_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Md8i71zKNBC7auihtad2mS"}]}]}} {"custom_id": "Rangpur_17955_21033_18468_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KPgTqt5GpxoSugyrzWN9z6"}]}]}} {"custom_id": "Gaibandha_20007_4104_20520_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KL9v8QFzSMsrVeMEWSSPh2"}]}]}} {"custom_id": "Jessore_16929_23598_17442_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WriiAMekog2prw8Yz7HSnD"}]}]}} {"custom_id": "Thakurgaon_8721_9234_9234_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XvP2EPNtKuLKzajhvn6BQ5"}]}]}} {"custom_id": "Satkhira_7182_34884_7695_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-G5xsuzXwuw4RgwwPG4ug5V"}]}]}} {"custom_id": "Bogra_17955_18468_18468_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Uwf61ai6XmZrpjEAJsG5th"}]}]}} {"custom_id": "Nawabganj_2565_15390_3078_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-85Cfh2T8Qd2rBhPcRSYB3q"}]}]}} {"custom_id": "Khulna_14877_11286_15390_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AfkVYqptLijSAgKdcmX8PM"}]}]}} {"custom_id": "Lakshmipur_13338_13851_13851_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-13B31Nv1MWUczD1pY3kiQD"}]}]}} {"custom_id": "Kishoreganj_23598_10773_24111_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R5fPefuZRCX1ovvmkHKJo4"}]}]}} {"custom_id": "Sylhet_4617_10773_5130_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JWb5JfhdXijUhLwrntRKK6"}]}]}} {"custom_id": "Sunamganj_14364_14364_14877_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2vKARdUy9UjdRDfaC4cgs3"}]}]}} {"custom_id": "Khulna_7182_42066_7695_42579.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9MZyaJddj2qT5SKZAd6jEe"}]}]}} {"custom_id": "Bandarban_16929_27189_17442_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JS6KXaXBUQ8QLAXBW5DyPw"}]}]}} {"custom_id": "Comilla_10260_3591_10773_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KDGmBQfdzzfXMzJpCq6xgC"}]}]}} {"custom_id": "Rajbari_8208_14877_8721_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6nXMyRVSCKqkcgy9Y7P7dD"}]}]}} {"custom_id": "Chuadanga_8721_17442_9234_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7Fd5wZqVYsSGAyybu6KmXK"}]}]}} {"custom_id": "Khulna_9747_31806_10260_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S6875n8pc1mbCkuLgKXVzj"}]}]}} {"custom_id": "CoxsBazar_6669_2565_7182_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3ziVndXiz8841ZuHGRHbDn"}]}]}} {"custom_id": "Rangpur_20520_7695_21033_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3ywE3JicahGyxU4QzZmd6q"}]}]}} {"custom_id": "Natore_18468_12312_18981_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KTG2ZimmS34VoJ4EKKwwvW"}]}]}} {"custom_id": "Rangpur_22572_24624_23085_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LNnk4wapK5mK3pNq6V9LnD"}]}]}} {"custom_id": "Sylhet_21033_12825_21546_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GjQPnyz8X9CswZuYAn687Q"}]}]}} {"custom_id": "Bandarban_17442_29241_17955_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VmtKRttTi6WJLGXqUKiQrC"}]}]}} {"custom_id": "Jamalpur_7695_30267_8208_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W1v8WnBPucEhpmi3GLujk8"}]}]}} {"custom_id": "Gazipur_10260_5643_10773_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1S9zn3DgjYHQRGYLgKt7xc"}]}]}} {"custom_id": "Thakurgaon_4617_10260_5130_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X2b3yVRvGtYzvFJcbzbm1G"}]}]}} {"custom_id": "Lakshmipur_3078_12312_3591_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JCoUwVNhGdsNnvRC4VjNry"}]}]}} {"custom_id": "Khulna_3078_14364_3591_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S5tTMLhKM3QBn2DpWDb9BW"}]}]}} {"custom_id": "CoxsBazar_8721_14364_9234_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JErZL5qneaStQAvrSpv3xY"}]}]}} {"custom_id": "Chittagong_32832_31806_33345_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FNqggRi83EmKmDKP1yUgqt"}]}]}} {"custom_id": "Pabna_4617_13851_5130_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HbWKw1rUbWRhGBZjDp9MJU"}]}]}} {"custom_id": "Sunamganj_5130_15390_5643_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XPHo1xtcbQuFTGtqQ3Y4md"}]}]}} {"custom_id": "Gaibandha_16929_22572_17442_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3heEjhwGw1xC1cSoCtkyvh"}]}]}} {"custom_id": "Chittagong_33345_27702_33858_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1rVEaCh1hCj4L7vf6VhoAd"}]}]}} {"custom_id": "Sylhet_13851_25650_14364_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-92PU93q5R5EabsqHZJ2v7F"}]}]}} {"custom_id": "Feni_6156_17442_6669_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1joXekHHipeA4GgNuLvWpq"}]}]}} {"custom_id": "Gopalganj_2565_9747_3078_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8HyUByiWvjrgiiqUuzmHQS"}]}]}} {"custom_id": "Joypurhat_5130_18981_5643_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WbgTVQ1qPsc9RmhPL9dMwJ"}]}]}} {"custom_id": "Comilla_20007_26163_20520_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8Yh1SMp19RAGJpypGbJuKT"}]}]}} {"custom_id": "Jessore_16929_10260_17442_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QsQYrnyoKxpzKmwsj3iSVz"}]}]}} {"custom_id": "Bagerhat_6156_22572_6669_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5Hww9ZqohPQ9Kt4uvCQjCd"}]}]}} {"custom_id": "Netrakona_13851_29241_14364_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-36XstgBAyt8m6UQLtjsYu2"}]}]}} {"custom_id": "Rangamati_31806_94905_32319_95418.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2QHUfGsj9QvERQeePgeNUW"}]}]}} {"custom_id": "Rajshahi_2052_10260_2565_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9gnMkWg1pYoTYLTYQvgxHN"}]}]}} {"custom_id": "Rangamati_17442_29754_17955_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K6PP7ZjtPV9NeTW4kh3TS4"}]}]}} {"custom_id": "Joypurhat_6156_6156_6669_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AHdiU164nhx5LvfoGiA2Ew"}]}]}} {"custom_id": "Jamalpur_11799_16416_12312_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LdDehPpeqZYDg91ymtDE2a"}]}]}} {"custom_id": "Khulna_9747_41040_10260_41553.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FRZ7xkAMsPGaETD4LkixFC"}]}]}} {"custom_id": "Narsingdi_4617_10260_5130_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JcQzQs4yXyia11BXkP87GQ"}]}]}} {"custom_id": "Gopalganj_8208_10773_8721_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3Rf7jjKvxkgCgKtSvLZRcd"}]}]}} {"custom_id": "Kishoreganj_11286_10260_11799_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4r3KBJEJAywG3gpXjnqY3F"}]}]}} {"custom_id": "Bagerhat_17955_14364_18468_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WbUdrF9831bJyvMX8K4nYk"}]}]}} {"custom_id": "Chittagong_29241_42579_29754_43092.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7oXKBudoBFw6HNntjzNy1b"}]}]}} {"custom_id": "Narayanganj_5130_14364_5643_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DRyonjgucfkjYWyMY2n6NY"}]}]}} {"custom_id": "Chandpur_15390_12825_15903_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Srj4eDycw3wZd2fDo45qn8"}]}]}} {"custom_id": "Nilphamari_11286_7695_11799_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EburYfcbCEh3HpdQnu8z1i"}]}]}} {"custom_id": "Satkhira_16929_38988_17442_39501.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RxiECpFQWWvmwMtSPy9pKA"}]}]}} {"custom_id": "Jessore_10773_8721_11286_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4k5UfeU76V5rNPoNXAHxvU"}]}]}} {"custom_id": "Shariatpur_11799_6669_12312_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DVKsvhKVLovzjseLxN7BzV"}]}]}} {"custom_id": "Thakurgaon_7695_1026_8208_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9qdWKd5CmLV6gc7uShKrjz"}]}]}} {"custom_id": "Thakurgaon_6669_11799_7182_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3x35CXTX9ijNWnUocwAmUg"}]}]}} {"custom_id": "Rangpur_10773_2565_11286_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Xtwgq75ap139gMcAV6U6nB"}]}]}} {"custom_id": "Gopalganj_8208_13851_8721_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4EdkrbnDvFpi6Ac3y7SqaP"}]}]}} {"custom_id": "Naogaon_28728_23085_29241_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3YRudtdeZKufxC3Tj9Mtdd"}]}]}} {"custom_id": "Gaibandha_9234_10773_9747_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SStrR6LLimbaAL822jxj7w"}]}]}} {"custom_id": "Maulvibazar_20007_22572_20520_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VP5zYywpqTzpT65Fwfja68"}]}]}} {"custom_id": "Lalmonirhat_13338_20520_13851_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D9NDDNBFCLnFjJ3XbULKgg"}]}]}} {"custom_id": "Manikganj_2052_9234_2565_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MN8gu8gG1sLeUqL4BgH8Ja"}]}]}} {"custom_id": "Jessore_12825_5130_13338_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VYAsWeHG2s7jAdkZB2HnUf"}]}]}} {"custom_id": "Patuakhali_21546_6669_22059_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1qryNUTiot67M9bGDGhrUa"}]}]}} {"custom_id": "Thakurgaon_13338_3078_13851_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hk2rWCCJ9eoWCnyJmtbt8v"}]}]}} {"custom_id": "Dinajpur_11286_25650_11799_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NHUJmAi9WZFoB53y6K1Yz7"}]}]}} {"custom_id": "Dinajpur_12825_16929_13338_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SGqi2DHSncbmXEDUDU5WYS"}]}]}} {"custom_id": "Khagrachhari_15903_16929_16416_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AgiqgJdibazyJ6W2Xx7Uuu"}]}]}} {"custom_id": "Bhola_4104_2052_4617_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2uMhbHXXaxwF4vkh3sBzNB"}]}]}} {"custom_id": "Panchagarh_513_7695_1026_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XWp3XKES8vcurU92CYnVHc"}]}]}} {"custom_id": "Naogaon_22059_10260_22572_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JYXBeM8NHJ9ugsQ5cpfWGZ"}]}]}} {"custom_id": "Nawabganj_6669_6669_7182_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AyXNyHBadTv6b7ewG7JvaG"}]}]}} {"custom_id": "Shariatpur_7695_4617_8208_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LMtngnwfwUitTEyKMaHZD8"}]}]}} {"custom_id": "Natore_21033_9234_21546_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5eGc3TiwhkZviygCW9V6oW"}]}]}} {"custom_id": "CoxsBazar_12312_39501_12825_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HjUAVaBHS3kEQKBv7VjN9g"}]}]}} {"custom_id": "Kushtia_5643_10773_6156_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QH2RPwcYSN28aZSE3nb4E2"}]}]}} {"custom_id": "Maulvibazar_3591_13851_4104_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VzGQnBEwYwxusQd3faVxxm"}]}]}} {"custom_id": "Maulvibazar_5643_32832_6156_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MtGjrEH8D1bGf6G3PfrgjM"}]}]}} {"custom_id": "Patuakhali_17442_36936_17955_37449.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FEPh7MXUZ9Y7gQyn6XL2Lf"}]}]}} {"custom_id": "Sirajganj_16929_31806_17442_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EszNnPjng3S6Q8UH8T4p2W"}]}]}} {"custom_id": "Kurigram_15903_28215_16416_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-19RnVU3fhTGHSjKX1wNrLA"}]}]}} {"custom_id": "Feni_3591_12825_4104_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4z4REUPDRbGszLZx1ZtN4A"}]}]}} {"custom_id": "Dinajpur_26676_33345_27189_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Dr1YxWUv3gjFnQQvDPE5kP"}]}]}} {"custom_id": "Jessore_2565_13851_3078_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-18H7iiCjtCeEspzFi3UsD7"}]}]}} {"custom_id": "Netrakona_11799_24624_12312_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UUR8YM99D6ZpWBRtPqGq88"}]}]}} {"custom_id": "Narsingdi_8721_2052_9234_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CVoJCRLQbRuMJF7PoxdHNm"}]}]}} {"custom_id": "Kishoreganj_14877_22572_15390_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MuVPxe5m1M8WDww5HfrF6G"}]}]}} {"custom_id": "Faridpur_22059_9747_22572_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PQqfbzZt9tGisEgtrQNcs7"}]}]}} {"custom_id": "Nawabganj_9747_9234_10260_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X2WRZPNKzn4PAV5caXbLEY"}]}]}} {"custom_id": "CoxsBazar_7182_17442_7695_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-81fVRtxkkq6FmrzyPT8SCY"}]}]}} {"custom_id": "Rajshahi_25650_11286_26163_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PaUswbtqv7Yby5ZZm8vmRp"}]}]}} {"custom_id": "Natore_15390_7695_15903_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5GeeRsUXKjbez9U69w35CL"}]}]}} {"custom_id": "Gopalganj_15903_18468_16416_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SGTPzQsWpBBkKFgz3zYwew"}]}]}} {"custom_id": "CoxsBazar_4617_16929_5130_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1k7aL3SqgVLZd4RM3N9wbM"}]}]}} {"custom_id": "Barguna_14877_2565_15390_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7WMFe1zMSrBwSgbtfe6BDq"}]}]}} {"custom_id": "Maulvibazar_2052_23085_2565_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3fSJTxeyZBjxikwbKUwGXf"}]}]}} {"custom_id": "Sirajganj_18468_31806_18981_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PdqyTnmze6WiGtztPRmvfz"}]}]}} {"custom_id": "Noakhali_16416_21033_16929_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QgUCnRY2PrC8a3K7YTdQpJ"}]}]}} {"custom_id": "Tangail_15390_22572_15903_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7wZ6WMV1uLnat1NgLxPdV8"}]}]}} {"custom_id": "Satkhira_13338_42579_13851_43092.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3iopktHT9GZ5Mqr1HEVd8g"}]}]}} {"custom_id": "Pabna_13338_6669_13851_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KUCwesyKdHQCdR1x8AqeN3"}]}]}} {"custom_id": "Mymensingh_20007_20520_20520_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6ogUw8N4zrZhMdAveQCEZa"}]}]}} {"custom_id": "Comilla_13338_14877_13851_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FYfQmPoepJuPjThVaDkG32"}]}]}} {"custom_id": "Naogaon_0_20007_513_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3vTHJaoNq2FJVBcLc8irxZ"}]}]}} {"custom_id": "CoxsBazar_12825_14877_13338_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Jfa1QdoKcPNKZD2r9P2qXZ"}]}]}} {"custom_id": "Sirajganj_18468_33345_18981_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9FY5Nxi2oooRnvyxyYkaGG"}]}]}} {"custom_id": "Khulna_14364_46683_14877_47196.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PjAA6wLgWBFpBvByog9K4b"}]}]}} {"custom_id": "Chittagong_27189_17442_27702_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Co3EufjcJq7xcWFdiLnQR7"}]}]}} {"custom_id": "Kurigram_13338_38988_13851_39501.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MfM3GPxGyxd1bsysRKUWtn"}]}]}} {"custom_id": "Brahamanbaria_18981_27189_19494_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CtQDzvLpHeh9yyEYCirpXX"}]}]}} {"custom_id": "Sirajganj_20520_20520_21033_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XHTJ8z3D8Lz5bGCcRYnxGk"}]}]}} {"custom_id": "Sherpur_2565_1539_3078_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P6sHR5QsQEawkDAu1TCZ9S"}]}]}} {"custom_id": "Tangail_3078_12825_3591_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T162Mabbgr5EMTdxWEXpMB"}]}]}} {"custom_id": "Kushtia_19494_17442_20007_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JF5DuJq2Vu4NbWktuxCJas"}]}]}} {"custom_id": "Rangamati_21033_42066_21546_42579.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J6nkg9qDfWTXMDZWQEKr8w"}]}]}} {"custom_id": "Bogra_19494_5643_20007_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HmAdAvXmVZ2BVptLZYMZtf"}]}]}} {"custom_id": "Narail_6156_11286_6669_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Q7xBM4Vv7xGYodozYyV38s"}]}]}} {"custom_id": "Bhola_20007_21033_20520_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4NYsMM7YkaphRc4SpHcUEL"}]}]}} {"custom_id": "Bhola_3591_46683_4104_47196.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MVzDqetTbkWc75VDhtzAFF"}]}]}} {"custom_id": "Noakhali_2052_2565_2565_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QwYpF5pwBWfeWYp3k22ChH"}]}]}} {"custom_id": "Rajshahi_10260_18981_10773_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SDng5LCsYJtkoAT2875QP1"}]}]}} {"custom_id": "Nawabganj_13851_24624_14364_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3KXKmAUzC6ftGHZ5MpAnSx"}]}]}} {"custom_id": "Bogra_27189_13338_27702_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3ShVcLK7gKpnGqs8cAHiNb"}]}]}} {"custom_id": "Rajshahi_7182_5130_7695_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TF6389BETRSr3WfJiumDsx"}]}]}} {"custom_id": "Kishoreganj_29754_4104_30267_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8csuwjyvPVkLQjDCFCSBTf"}]}]}} {"custom_id": "Bogra_18468_14364_18981_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8ChFnWvaQHKej4M4d2Nn6r"}]}]}} {"custom_id": "Sylhet_5643_0_6156_513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7pkftHGRmcQbuq9giu1Mdj"}]}]}} {"custom_id": "Faridpur_12825_16416_13338_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RLY1PHbD5DuMCSUdmLazgk"}]}]}} {"custom_id": "Bhola_6156_2565_6669_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3Wb6mzjyuKjHCXDSp8vbF8"}]}]}} {"custom_id": "Tangail_9234_16929_9747_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ADu1mFTd1cMGdJVuU8PsDm"}]}]}} {"custom_id": "Feni_10260_20520_10773_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FSB6u322VZB24FrRy2D14q"}]}]}} {"custom_id": "Comilla_24111_10260_24624_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3WyvycaXC4ULzXWkFmZWHw"}]}]}} {"custom_id": "Chittagong_28215_19494_28728_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-45L6hkztNSZULTMXCFZRsQ"}]}]}} {"custom_id": "Narsingdi_8208_3591_8721_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9tkhj3Yt88PrFbknMVkjNR"}]}]}} {"custom_id": "Jessore_25650_16929_26163_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6N7hTmHEX1wtKMg37nAAS9"}]}]}} {"custom_id": "Bandarban_6156_53865_6669_54378.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FPFMMDqHZLZYj5RUqRsgTF"}]}]}} {"custom_id": "Sirajganj_20520_13851_21033_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BcnRxLqmoewppgJ483kx1H"}]}]}} {"custom_id": "Shariatpur_15390_11799_15903_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SWfNH8qDkYvuLhvTQjqZxr"}]}]}} {"custom_id": "Shariatpur_12312_9747_12825_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1L9y6dbnE4boj6h9gcEVTN"}]}]}} {"custom_id": "Sirajganj_14364_26163_14877_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JBBMsNFHf264tDP3jZREMh"}]}]}} {"custom_id": "Mymensingh_18981_45144_19494_45657.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QmDthx4RuwSp8Emiioyek8"}]}]}} {"custom_id": "Narayanganj_1026_12312_1539_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Cd4WJPhCaReiMfycm1zPE8"}]}]}} {"custom_id": "Sunamganj_24624_13338_25137_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ai6tjvwarRNkRnGoLGw8fv"}]}]}} {"custom_id": "Dinajpur_6669_6669_7182_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UeLTBsYqESHz4fraw6kiak"}]}]}} {"custom_id": "Sylhet_3591_29241_4104_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RDriHmdwPPpxBufgKq1q6Y"}]}]}} {"custom_id": "Rajbari_11799_15390_12312_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BrjvX8CT814Sa88KaDiPXv"}]}]}} {"custom_id": "Rangpur_17955_28728_18468_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3KQsQVwVoP5aqi5PRzJqLU"}]}]}} {"custom_id": "Khagrachhari_3078_36936_3591_37449.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DBUe2JVnDVtaVi4nRF8bVm"}]}]}} {"custom_id": "Meherpur_11799_7695_12312_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TxLdwwGi7ghDyn1RC13NGk"}]}]}} {"custom_id": "Gazipur_8208_2052_8721_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VG8qN5SwwAtFMCMsKBCh6T"}]}]}} {"custom_id": "Narsingdi_9234_19494_9747_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ex54XAAmHs5Y3Wa57Ds7uQ"}]}]}} {"custom_id": "Natore_16929_22572_17442_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A1oDTkKyQwYBHU5wizoZEg"}]}]}} {"custom_id": "Jhenaidah_24624_3078_25137_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TivQ4bP573a4HkoD5d6Mtf"}]}]}} {"custom_id": "Kurigram_5643_32832_6156_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AAwe6TfTjicPJ114NqtkTG"}]}]}} {"custom_id": "Sunamganj_18468_9234_18981_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L4VEV9B18UNcuwnurRFrzt"}]}]}} {"custom_id": "Chandpur_3591_10260_4104_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6RUVimStZEP6RL8kdtuocA"}]}]}} {"custom_id": "Mymensingh_34371_32319_34775_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EeUiTvCQjwpcCXdm2ijTBX"}]}]}} {"custom_id": "Bagerhat_13851_53865_14364_54378.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7an2JDjvYM1WNwcAqXprr9"}]}]}} {"custom_id": "Kishoreganj_18981_2052_19494_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ADpxgzRqLF6j6sNkZGjHGT"}]}]}} {"custom_id": "Jessore_18981_11286_19494_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CcMkMD7j2wWDwwu5sBv9az"}]}]}} {"custom_id": "Khulna_16416_2052_16929_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Y9Xf6Mg6ktVixKHw48nQmz"}]}]}} {"custom_id": "Jhenaidah_17955_4617_18468_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TRGxU56EpeBGwMscMuRiYj"}]}]}} {"custom_id": "Rajbari_9747_15390_10260_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TxBAr76Jnnb74iSXEvdkyp"}]}]}} {"custom_id": "Maulvibazar_18468_15390_18981_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WQ7ZkoV6EpnyEwmhg6A7Vn"}]}]}} {"custom_id": "Munshiganj_12312_4104_12825_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4XwbLEh6v6YQAKbVX35nLR"}]}]}} {"custom_id": "Naogaon_3078_22572_3591_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N5cgaQ9MqRxenvfHjZHd9L"}]}]}} {"custom_id": "Khulna_9234_14877_9747_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ER9kSHHEDXvLiJFUMVS15Q"}]}]}} {"custom_id": "Bandarban_21033_46683_21546_47196.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mq4g4yd1wGRDs4JyGvQiZ2"}]}]}} {"custom_id": "Maulvibazar_9747_16416_10260_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FQugSt8J6CV1yJtHUoczBw"}]}]}} {"custom_id": "Pabna_11286_17955_11799_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9ThEeSckkqgxzj5udVR1LK"}]}]}} {"custom_id": "Sylhet_17442_24111_17955_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NFxsfgbNNLyGhYSo61rqo8"}]}]}} {"custom_id": "Jhenaidah_8208_18981_8721_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XooLdyfjyZnNYeCaEJPCji"}]}]}} {"custom_id": "Bandarban_9747_9234_10260_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L3wxqNJdKwCNAfWFynsgaf"}]}]}} {"custom_id": "Nawabganj_9234_18981_9747_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RC6mH1yicCTpNfBgr1yj1h"}]}]}} {"custom_id": "Patuakhali_1539_37449_2052_37962.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GgZCuaEARGsitPRgdMni8M"}]}]}} {"custom_id": "Mymensingh_24624_23598_25137_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CpLoXbAUDCBZvDpaC8iStZ"}]}]}} {"custom_id": "Rajshahi_18981_6156_19494_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3g1Z55yJThDufxK7FdBE66"}]}]}} {"custom_id": "Comilla_15903_12312_16416_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3YtGVtcsYagbyR4WncGREH"}]}]}} {"custom_id": "Pabna_21546_16929_22059_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XPTAddDDj51FZTmjeqqQX6"}]}]}} {"custom_id": "Satkhira_16416_27702_16929_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VqAb9P7pZ3ySM7g1mEgmyw"}]}]}} {"custom_id": "Chuadanga_8721_12825_9234_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-83S1XrEWUADLBJa9asxmVB"}]}]}} {"custom_id": "Kishoreganj_20007_15390_20520_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-12hCwVT1kCfcvSd2m26JZW"}]}]}} {"custom_id": "Shariatpur_11286_17442_11799_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2kEM6H6RcdS37edfVhyNbK"}]}]}} {"custom_id": "Thakurgaon_4104_13338_4617_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GgQG8vnWT6evQVUMMnyZXw"}]}]}} {"custom_id": "Gopalganj_13851_7695_14364_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JQGSwao2iETjbffHhkQyua"}]}]}} {"custom_id": "Chittagong_34884_23598_35397_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5B8bGqjhrHNpBBUVcvvT1n"}]}]}} {"custom_id": "Mymensingh_9234_11286_9747_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-81ahDgdBjpbUTuyk7pwveM"}]}]}} {"custom_id": "Sylhet_11286_3078_11799_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3TLMT1cEbzryKgeSFTpp8r"}]}]}} {"custom_id": "Barguna_18468_20520_18981_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FwT3yi2JoZKo834jzeDLgp"}]}]}} {"custom_id": "Tangail_14364_2052_14877_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-45NbSRSMoRB7aSu1sDTcd6"}]}]}} {"custom_id": "Narayanganj_7182_10260_7695_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TnoeVDrBdHMXwX1ULBVSWc"}]}]}} {"custom_id": "Jessore_18981_23598_19494_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XeqwvZxSq3ZPHhMRui9WdE"}]}]}} {"custom_id": "Madaripur_11286_15903_11799_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ApxW9SkeHzGFWYbkEDUcc3"}]}]}} {"custom_id": "Gazipur_15390_7695_15903_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bd1m7TFtn7665TKvj7vFLM"}]}]}} {"custom_id": "Munshiganj_9234_5130_9747_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GvT3j7THNHMLqanrTgeCcV"}]}]}} {"custom_id": "Jamalpur_8208_22059_8721_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MK5sipbdPKJdFezHMywpiL"}]}]}} {"custom_id": "Shariatpur_4617_4104_5130_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fqy4uxWW9cmma87AuqqfWo"}]}]}} {"custom_id": "Shariatpur_2565_9747_3078_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8ubr4fcb8iXuxFTysfhpU5"}]}]}} {"custom_id": "Chittagong_12312_8721_12825_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CGnBVcSgzD3ufuxaLh45up"}]}]}} {"custom_id": "Naogaon_27702_17442_28215_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1qJYwo4jBj5tmiCSZSa36X"}]}]}} {"custom_id": "Sylhet_5130_22572_5643_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2v9CgHGW3mR76rRPG7FVVS"}]}]}} {"custom_id": "CoxsBazar_20007_46170_20520_46683.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LjfRNjcG45p32uzge1xWXE"}]}]}} {"custom_id": "Dinajpur_21546_31293_22059_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-64jijCFViwYfvN4h7uyfw4"}]}]}} {"custom_id": "Lakshmipur_10260_1026_10773_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Uf2BEaRdGCP2tyQftYyTkn"}]}]}} {"custom_id": "Tangail_18468_12825_18981_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1k6itodD2ttqYXQeGRzQGF"}]}]}} {"custom_id": "Jessore_29241_16929_29754_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W53FsaxMpx6iW5QbviK8kM"}]}]}} {"custom_id": "Bhola_9747_14877_10260_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6Lz2Kd1ZateTc8ALrj5M2M"}]}]}} {"custom_id": "Habiganj_14364_27702_14877_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MGp46rDnqWPMNGBFWxHBvt"}]}]}} {"custom_id": "Manikganj_10260_11799_10773_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BBws1JWnqTwffaxkUdq6st"}]}]}} {"custom_id": "Lalmonirhat_23598_22572_24111_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mo7g4f42hA2dEkDtRv3GAV"}]}]}} {"custom_id": "Magura_10773_3078_11286_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5Yzf5BNsvGUDTnBGHs2GaT"}]}]}} {"custom_id": "Naogaon_14877_11799_15390_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bpqi87WGitNpjNWFgdBgz9"}]}]}} {"custom_id": "Shariatpur_9234_10773_9747_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GRY31bppXQacbyjtbKex6Z"}]}]}} {"custom_id": "Gopalganj_10260_2565_10773_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NgSXoTZeesEV3zKamgSUJb"}]}]}} {"custom_id": "Brahamanbaria_17955_18468_18468_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PHTkVBnXF8L5g83H1dfPd8"}]}]}} {"custom_id": "Barisal_15903_22059_16416_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UuHFcBLhWxLqGHxeSjZvKV"}]}]}} {"custom_id": "Bagerhat_10260_7182_10773_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TjT14FUTycpwvLkRmeX1aF"}]}]}} {"custom_id": "Feni_13338_11799_13851_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S8ma94Zah8UHanVjj6WcDG"}]}]}} {"custom_id": "Patuakhali_8721_33858_9234_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PBkRbbr9EhEpr72oHVBEgj"}]}]}} {"custom_id": "Lakshmipur_0_8721_513_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AC7C8qAEnnTxy4cLeCG31Y"}]}]}} {"custom_id": "Rajshahi_19494_18981_20007_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HTs8TX4xTRMVp5eYY2ifJg"}]}]}} {"custom_id": "Feni_2565_22059_3078_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-92MDNceR9jU15sS8BxipkE"}]}]}} {"custom_id": "Jamalpur_2052_12825_2565_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hg7gX37GQc9nDm8ZqxUjqx"}]}]}} {"custom_id": "Rangpur_13338_20520_13851_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YTnetgtS5um9Q6tUaPPeTL"}]}]}} {"custom_id": "Panchagarh_21546_24624_22059_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DkRG8frAnNj6G9RHVDAyWd"}]}]}} {"custom_id": "Barisal_20007_8208_20520_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U5qy4WpdfCqpxn564YkHeB"}]}]}} {"custom_id": "Sylhet_29754_8721_30267_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UHrNFnxhZT1wsaL5jYg4HS"}]}]}} {"custom_id": "Kishoreganj_22059_28728_22572_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UCkbxvryS4t7Y8PvdHgbQJ"}]}]}} {"custom_id": "Patuakhali_12825_21033_13338_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8xPFo6csMXrBGrMRqWCraP"}]}]}} {"custom_id": "Pirojpur_3591_24624_4104_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XYqGwgLc93PaUaZ4EhMJ3g"}]}]}} {"custom_id": "Noakhali_4617_34371_5130_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RJN37fjae1sRorZZko5p1D"}]}]}} {"custom_id": "Satkhira_12825_32319_13338_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2MM7d64RoHqLNHdmVBZmky"}]}]}} {"custom_id": "Chittagong_14877_15390_15390_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2kHt28RKyNrYByurQhxbf7"}]}]}} {"custom_id": "Comilla_10773_17442_11286_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9bAQRA9dU2DaMytyKCDj5x"}]}]}} {"custom_id": "Chuadanga_7182_19494_7695_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FfGLR9i5W9dQubeQv7U5pm"}]}]}} {"custom_id": "Rangamati_15903_36423_16416_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-G9pGFu2Xh7bbhxR7rBUA3N"}]}]}} {"custom_id": "Pabna_13851_15903_14364_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DsR6gFvnTZ118uhRqr3ZnN"}]}]}} {"custom_id": "Pirojpur_6669_10260_7182_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-32SUecG5vUCQ1y7hgAhicv"}]}]}} {"custom_id": "Munshiganj_6669_3078_7182_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CNNyaMLtsKW88k1b855EAN"}]}]}} {"custom_id": "Thakurgaon_15390_12825_15903_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KoDWAQhdvBMUbnS4izLqjD"}]}]}} {"custom_id": "Pirojpur_3591_4104_4104_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K5mtZa4gTBAvvs92oum8vS"}]}]}} {"custom_id": "Rangamati_9747_66690_10260_67203.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2YySub4EhXj6Pbj5Dxq37t"}]}]}} {"custom_id": "Barisal_23085_15390_23598_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N5142H2nC8SsU9XwBm5iP3"}]}]}} {"custom_id": "Munshiganj_24624_9234_25137_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gjj4jJ4KvkUaKGXZUybZXT"}]}]}} {"custom_id": "Dinajpur_35397_38475_35910_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5gfwDS6M5puxzWBB6stNto"}]}]}} {"custom_id": "Comilla_24111_13338_24624_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ETL8Avm51RRwg8RoPc832U"}]}]}} {"custom_id": "Tangail_14364_15903_14877_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Baqp81XMsv4W6tSbTbdnRs"}]}]}} {"custom_id": "Tangail_8721_10773_9234_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8i76gf9VSMo9Zami2xvmg5"}]}]}} {"custom_id": "Manikganj_12312_8208_12825_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XZDJTXfJcwfr5rtxnqnQqb"}]}]}} {"custom_id": "Rajbari_5130_2565_5643_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EHPkycey2zxgeWA4jJrKKm"}]}]}} {"custom_id": "Feni_8721_17955_9234_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HsG8Uz67tUYnkDqLFr7Loa"}]}]}} {"custom_id": "Khulna_10260_5643_10773_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P3HNA1oy6meNTMYjoZPYXa"}]}]}} {"custom_id": "Chuadanga_7695_16929_8208_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8hK5Rzs1gFvFDYhZ3Sv3LQ"}]}]}} {"custom_id": "Kurigram_5130_29754_5643_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2u2k7qRCnyhK2P8E4BeM55"}]}]}} {"custom_id": "Mymensingh_5130_28215_5643_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Afarqf8q9G22x7hMpY8CYg"}]}]}} {"custom_id": "Naogaon_14877_4617_15390_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DN2PvMHS511tmS4jznTK3B"}]}]}} {"custom_id": "Natore_14877_24624_15390_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YPHVZGp926z74nXPwMKKJb"}]}]}} {"custom_id": "Tangail_14364_5643_14877_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SrPsqhhb7bk5tec7qTNdRM"}]}]}} {"custom_id": "Bandarban_16929_3078_17442_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R4QSApk56kaotXyLK9Ezno"}]}]}} {"custom_id": "Nilphamari_4617_17442_5130_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ViuyhF3dFJXouBnMsHuTZW"}]}]}} {"custom_id": "Brahamanbaria_21033_10773_21546_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WdyBh7dkWjUNs4rtmjabGA"}]}]}} {"custom_id": "Naogaon_23598_2565_24111_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DcBj4wmUA1Tknxh8iBexHE"}]}]}} {"custom_id": "Gazipur_15903_11799_16416_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VcBGhYbFWpFXNy2EERqhKn"}]}]}} {"custom_id": "Tangail_15390_25650_15903_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8eLHFBa2E8VguxmWGdGWHq"}]}]}} {"custom_id": "Chuadanga_6156_11286_6669_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EdySfwx633P4Ps4PpdLi5i"}]}]}} {"custom_id": "Madaripur_5643_14877_6156_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BNEscyX2ku2iNhKtjfea26"}]}]}} {"custom_id": "Mymensingh_16929_34884_17442_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7NKtxi1dQ4UCvXXyLxHSph"}]}]}} {"custom_id": "Maulvibazar_21033_9747_21546_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CF3P4cVTarpVReDoJruGV4"}]}]}} {"custom_id": "Dinajpur_28728_21033_29241_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HoP6ZiQRLUChnzeTyRtWS3"}]}]}} {"custom_id": "Narsingdi_11286_3591_11799_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3DdyxaZbqmqtvFKWDUCcGm"}]}]}} {"custom_id": "Jessore_8721_21033_9234_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8Fn93jPohW8ZkJxEHD5QxT"}]}]}} {"custom_id": "Tangail_16416_31806_16929_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-B5ccNYBvYNRg7fmQNPF4vM"}]}]}} {"custom_id": "Satkhira_14364_37449_14877_37962.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JtAmeQy3aac6J7bHKvjP2C"}]}]}} {"custom_id": "Bogra_15390_21033_15903_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VGxVaDNc6fMoK3384gEN4B"}]}]}} {"custom_id": "Sherpur_12312_20007_12825_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QjufRfm6HyW2bW9adVPLHE"}]}]}} {"custom_id": "Shariatpur_6669_9234_7182_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5jCALrAZ19LyUQcPZeNzMt"}]}]}} {"custom_id": "Sunamganj_34884_4617_35397_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K5vRAY8eU5mHBZAXLY4bqQ"}]}]}} {"custom_id": "Meherpur_11286_7695_11799_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3zE9RDsxYs3oMi3UT6gELx"}]}]}} {"custom_id": "Kishoreganj_14877_7695_15390_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3gND4pibZ4LtftxaGhSKay"}]}]}} {"custom_id": "Jamalpur_7182_20007_7695_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2JCdWdPeb3Gmimrpvrd8MJ"}]}]}} {"custom_id": "Sylhet_5130_9234_5643_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JvjKzfcCPdhjTmFXf9EDTD"}]}]}} {"custom_id": "Brahamanbaria_23085_26676_23598_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WCWjBjUcx1cLPtcxhVCFcK"}]}]}} {"custom_id": "Barisal_18468_17955_18981_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UyQ4bkPdjPff9Boiv4VWaY"}]}]}} {"custom_id": "Netrakona_14364_6156_14877_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CW93taicaXcnq66D8b8VqJ"}]}]}} {"custom_id": "Sirajganj_20520_35397_21033_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GbSbgLhuMfXe8mk3gWRWL2"}]}]}} {"custom_id": "Bhola_13338_42579_13851_43092.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NtN4HmPbMHuwbK5NrRraPF"}]}]}} {"custom_id": "Faridpur_14364_9747_14877_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6bUiv1fyxSpyfgSUpswuWX"}]}]}} {"custom_id": "Rangamati_23085_51300_23598_51813.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T423rjLeZS6SH4iiQHm72h"}]}]}} {"custom_id": "Gaibandha_8721_12825_9234_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hq7KNwk9skFkVgGYv3pFxv"}]}]}} {"custom_id": "Bagerhat_1539_20520_2052_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T7jbRSitaCn3ftwsqfEKoY"}]}]}} {"custom_id": "Sunamganj_27189_20520_27702_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MHNFBDsjkLC3nkDCs9Hwur"}]}]}} {"custom_id": "Rajshahi_10773_12825_11286_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-H6zs2CFwF2mowhJ1ZEVydg"}]}]}} {"custom_id": "Satkhira_10260_4617_10773_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LYgrXxgWcy5eRGqXCMXsra"}]}]}} {"custom_id": "Kishoreganj_14364_3078_14877_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ULG7whSyALHiNZ9we5gLgu"}]}]}} {"custom_id": "Gazipur_7182_4104_7695_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JFdpar6NJqvBWTto8TJrYW"}]}]}} {"custom_id": "Noakhali_12825_47196_13338_47709.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VH6xTscWNTki4a7N44Rgap"}]}]}} {"custom_id": "Tangail_7695_10773_8208_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JJEYjrTcTizNfh7hL459m7"}]}]}} {"custom_id": "Natore_3078_15390_3591_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WNAGK2YASMhqEJH2NNtopZ"}]}]}} {"custom_id": "Barguna_3591_23598_4104_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PFMJzBCUoNVo3YgDNxuVh9"}]}]}} {"custom_id": "Kishoreganj_12825_5130_13338_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8FsnRnZyX2C9ZiJtB8izC9"}]}]}} {"custom_id": "Satkhira_3591_6669_4104_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Eytv8evwGRTQB8eeMaRX19"}]}]}} {"custom_id": "Brahamanbaria_12312_14877_12825_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MDbAALKcafXGAZTm3fJLHj"}]}]}} {"custom_id": "Narsingdi_6156_4617_6669_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2aT71ykBwGRqr7o1VFsfTU"}]}]}} {"custom_id": "Chittagong_32832_51813_33345_52326.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EQvFLYd9n7cfZtViZiWYaL"}]}]}} {"custom_id": "Noakhali_10260_3591_10773_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PPHARsax87ajZyJPHiXuKv"}]}]}} {"custom_id": "Mymensingh_2052_24111_2565_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MEQDhdDYeEpXRf9o2UZDW5"}]}]}} {"custom_id": "Faridpur_21033_12825_21546_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JhsNy7xy9Mt83j2LE6r9Yy"}]}]}} {"custom_id": "Tangail_16929_29241_17442_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4mfgz4VCdbyyLqsYDicVWq"}]}]}} {"custom_id": "Tangail_14877_5643_15390_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XKKGoE1woLDmLjufPk9vno"}]}]}} {"custom_id": "Rangamati_6669_42066_7182_42579.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3N7aVU36QjjPM8tJzAKEbK"}]}]}} {"custom_id": "Barisal_15390_1539_15903_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QFKZUGhjDq4qAJeYRBUZLv"}]}]}} {"custom_id": "Sherpur_8721_19494_9234_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FrXK1aU7xY1N9UcGueYkDE"}]}]}} {"custom_id": "Maulvibazar_1026_26163_1539_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8kpoP1S8y8As1v6qmpbrmu"}]}]}} {"custom_id": "Sunamganj_12312_1026_12825_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CypHqcSga6opMzbYUHUSFj"}]}]}} {"custom_id": "Naogaon_6669_3078_7182_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QX83iUezAL37HUmgg1AtcP"}]}]}} {"custom_id": "Gazipur_11799_12825_12312_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YPntGKSDizN2rF3gSd3TBX"}]}]}} {"custom_id": "Kushtia_15903_6669_16416_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BNqYsS74sTKKDFTGS5zjLC"}]}]}} {"custom_id": "Rangamati_25650_49248_26163_49761.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FdwXJDvo87ks61fmzarYxX"}]}]}} {"custom_id": "Thakurgaon_18981_513_19494_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9kvoCgbmnzzxW96jeGeHYu"}]}]}} {"custom_id": "Chittagong_26676_50274_27189_50787.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JHVoQqW2RUeNeki8e6FBn9"}]}]}} {"custom_id": "Bagerhat_2565_56943_3078_57456.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AsDgMcRNxc7uuVifgeu74f"}]}]}} {"custom_id": "Brahamanbaria_1026_29241_1539_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ExsxqLXRCkAE8qQb3TEhSw"}]}]}} {"custom_id": "Maulvibazar_12312_27702_12825_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8u5eKZh1UNfpFESAbqdgSm"}]}]}} {"custom_id": "Sylhet_29754_13851_30267_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XT75TdPeWGCFgK5V2xEEJK"}]}]}} {"custom_id": "Thakurgaon_17955_8208_18468_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Du2MrPvk8tznGAQCo4f3xh"}]}]}} {"custom_id": "Barisal_14364_7182_14877_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7HMLZUUoTkRgWWh4Eznvmb"}]}]}} {"custom_id": "Rajshahi_25650_16416_26163_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GqMusSTKrxDZzND65oeSyF"}]}]}} {"custom_id": "Noakhali_14364_25650_14877_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FwrzeFaEjZp5SDpFjmf9Pw"}]}]}} {"custom_id": "Narail_8721_13851_9234_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FrxMWLtBZFN9emFZ4uQsV2"}]}]}} {"custom_id": "Netrakona_11286_5130_11799_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NKrg2mbjU6XN8sLChEM8gt"}]}]}} {"custom_id": "Bogra_24111_5130_24624_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CYV1ym9Uao2S3Ap8asgfk7"}]}]}} {"custom_id": "Khagrachhari_11286_44631_11799_45144.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LaXk5AHzW183ac298aWF9J"}]}]}} {"custom_id": "Sunamganj_2565_3078_3078_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-16wzEmstwSKuHVUsNMCEVR"}]}]}} {"custom_id": "Bogra_16929_12312_17442_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EgMrwS81pUwct25LbgmZ28"}]}]}} {"custom_id": "Jhalokati_10773_6156_11286_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FYaC7T8sED4S75sErYeqpW"}]}]}} {"custom_id": "CoxsBazar_12825_29241_13338_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9uW3GXPZiDsQtrcZ9znMEF"}]}]}} {"custom_id": "Thakurgaon_15390_2052_15903_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7fR4dSAPCqEE9xPcZhsVZw"}]}]}} {"custom_id": "Sylhet_2565_19494_3078_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EyPvuWo3e38VuZWuseKxYB"}]}]}} {"custom_id": "Mymensingh_14877_13338_15390_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KrSyTKFHF8Q5U9kmFLyaYd"}]}]}} {"custom_id": "Habiganj_15903_20520_16416_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SxCmLq2X6qz7RZJ8esdWy4"}]}]}} {"custom_id": "Mymensingh_11799_15390_12312_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VpTgAX27LvnbgB561Crnrn"}]}]}} {"custom_id": "Jessore_21546_11286_22059_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5DghqygosvB7sQfjux94xV"}]}]}} {"custom_id": "Pirojpur_4104_24111_4617_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XkG5TCE7HE3UEsEjMkkUCE"}]}]}} {"custom_id": "Thakurgaon_14877_10773_15390_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5FzB6dRPzEYZBjySUWpHw5"}]}]}} {"custom_id": "Dinajpur_28728_39501_29241_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ji3y9gdPq79iRJAw4FPiCU"}]}]}} {"custom_id": "Gazipur_7182_7695_7695_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SJWTdpWTDTPokJhGwj5bnC"}]}]}} {"custom_id": "Faridpur_28215_8721_28728_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4Z6TVH6Phg94Lu5koMJotU"}]}]}} {"custom_id": "Comilla_26163_29241_26676_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Jmv4QGYfi1G1VuNVK9umCw"}]}]}} {"custom_id": "Nawabganj_16929_4617_17442_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-46KneKT4bLRh3eyByN9ahw"}]}]}} {"custom_id": "Kishoreganj_29241_6669_29754_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gch1nqB6tU3tSPrzxyoytY"}]}]}} {"custom_id": "Jessore_1539_20007_2052_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9LGxwBUMcUKrd45V2JVXJ4"}]}]}} {"custom_id": "Tangail_6156_21033_6669_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TAyfJNYXSYYf1CtPBPPiZu"}]}]}} {"custom_id": "CoxsBazar_2052_12825_2565_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L9tHuXGQcfeaUxVjqNWmo6"}]}]}} {"custom_id": "Barisal_15903_1026_16416_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BnVn6T3pfRJBCfYw3E3sdT"}]}]}} {"custom_id": "Rangpur_16929_12825_17442_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FZ8J7qktLzDTwotyRrG6m9"}]}]}} {"custom_id": "CoxsBazar_7182_8721_7695_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VCJ5c3PwupRBjHdT3zesfA"}]}]}} {"custom_id": "Khulna_9747_43605_10260_44118.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XgreSqZdFmcXmy3ujZ6igN"}]}]}} {"custom_id": "Bogra_19494_3078_20007_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9u8i4H3D82zTffqBWpMDE7"}]}]}} {"custom_id": "Comilla_31293_31293_31806_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RcKU8TzTJD5B99CJoohdtb"}]}]}} {"custom_id": "Rangamati_25137_59508_25650_60021.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-26GJhYZhNRprMAL6ZTp8Qh"}]}]}} {"custom_id": "Faridpur_27189_18981_27702_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3ABKHv9b5QVmkDeHcvSahW"}]}]}} {"custom_id": "Barisal_6669_3591_7182_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DnR7cEXaPu9q4TBgatrgvi"}]}]}} {"custom_id": "Narail_5130_12312_5643_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F7d3mmTTePbU1uNYFSZjj5"}]}]}} {"custom_id": "Tangail_20007_28728_20520_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NxVLZKAHwd971GBfeg4Zf8"}]}]}} {"custom_id": "Bagerhat_3078_18981_3591_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LZGeuZf5ADDnUQwvbrP12t"}]}]}} {"custom_id": "Mymensingh_12825_38475_13338_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5atMpCt4mTD7t3suPKbK8P"}]}]}} {"custom_id": "Maulvibazar_2565_22572_3078_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VieDVksiQtqp9z9z1TjSmx"}]}]}} {"custom_id": "Comilla_11286_3078_11799_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TyDwhipNuz7XoNDyBGgyB3"}]}]}} {"custom_id": "CoxsBazar_20007_49761_20520_50274.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UsiSLRZ7vmeLBrJQrURXgR"}]}]}} {"custom_id": "Gazipur_3078_10260_3591_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VvdVmKNPU7ydQ7hVhbyFra"}]}]}} {"custom_id": "Lalmonirhat_18468_27189_18981_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DzNFGYuzyWv6KmbtmEA3o7"}]}]}} {"custom_id": "Jamalpur_16416_30780_16929_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fcy4YsnY3qRzdxBEtkgMSV"}]}]}} {"custom_id": "Sylhet_3591_25137_4104_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UiH7KLiT4VrLLyNoVjc5XL"}]}]}} {"custom_id": "Mymensingh_7695_10773_8208_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MTM5pVm2Z6NcgGDgqDb2ba"}]}]}} {"custom_id": "Patuakhali_12312_20520_12825_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8Cn6TajA2VtkFJJt7vDyyZ"}]}]}} {"custom_id": "Kishoreganj_22059_15390_22572_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Q4XNbWLH3Q8nBjmWBdtDY7"}]}]}} {"custom_id": "Rangamati_13338_32832_13851_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vr85joqvwkajbGwq5EW33q"}]}]}} {"custom_id": "Feni_5643_23598_6156_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KZy8KGasrnuUu9o74wQ2KA"}]}]}} {"custom_id": "Bagerhat_13338_4104_13851_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CXG8gFLNG9Jni17xWBPPRZ"}]}]}} {"custom_id": "Brahamanbaria_18468_15903_18981_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2T5NMHzZU3pKzH6SpVsZYM"}]}]}} {"custom_id": "Faridpur_12825_9234_13338_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P5zdoVXycqytd1GwfXz6cE"}]}]}} {"custom_id": "Jamalpur_11286_30780_11799_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1bcERzg2VGjZfbdQWQj3cY"}]}]}} {"custom_id": "Chittagong_20520_17955_21033_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HwTy9DGLVT3cPE4s6rmZRa"}]}]}} {"custom_id": "Bandarban_6669_25137_7182_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CpADYTgWYdiUBoJa3EXT4G"}]}]}} {"custom_id": "Bhola_3591_35910_4104_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3bN4KxpCmTokvwUNGYqJo8"}]}]}} {"custom_id": "Narail_3078_12825_3591_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2Yu8XMavmAMTQ12Uw3Sern"}]}]}} {"custom_id": "Narsingdi_14364_16929_14877_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S5sSTLP6CoiSGuWbiB8GtH"}]}]}} {"custom_id": "Kishoreganj_14877_24111_15390_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5dYtR8fPjed98coi4kYYAV"}]}]}} {"custom_id": "Naogaon_1026_19494_1539_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6cbha95zzkB99zVzkrB2Fq"}]}]}} {"custom_id": "Khulna_7695_27189_8208_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4zrFH1hG4NJ64bSsgPm7kg"}]}]}} {"custom_id": "Comilla_18981_28728_19494_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HfTu4hDjCX4MEBkuYeW14g"}]}]}} {"custom_id": "Satkhira_9234_33858_9747_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VtdWcgvL6xmuqqKUE42Twc"}]}]}} {"custom_id": "Patuakhali_17955_12312_18468_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ML5nfLe5b6uwdUVBjKAi8u"}]}]}} {"custom_id": "Sylhet_8721_15390_9234_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ps5KH8b3xVYgHQiXWFLo1J"}]}]}} {"custom_id": "Dinajpur_12825_22572_13338_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5bevBEhVaaJyhF6kdYsoQS"}]}]}} {"custom_id": "Noakhali_14364_8721_14877_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V2oZN8CtWp9SK83mHqF5Wk"}]}]}} {"custom_id": "Bagerhat_6669_53352_7182_53865.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ARQzzLFPhXYAwzo5ounUqt"}]}]}} {"custom_id": "Faridpur_19494_20520_20007_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XVJDYD8VaECYLmZHyyo5W7"}]}]}} {"custom_id": "Lakshmipur_8208_18981_8721_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JPGPLhMZsG9FQTem9UnYXA"}]}]}} {"custom_id": "Sunamganj_16416_15903_16929_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HFw24Fnz8BUGeJUWp3HknA"}]}]}} {"custom_id": "Bandarban_6669_14364_7182_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EMacbUQa1gaf6zgFmQfezE"}]}]}} {"custom_id": "Brahamanbaria_9234_22059_9747_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3mLC4RAy1teq24AH3NMpdK"}]}]}} {"custom_id": "Maulvibazar_15903_10773_16416_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TDwwe4waG7PwXe774NkEhX"}]}]}} {"custom_id": "Bagerhat_16416_7182_16929_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JMqRgYjXnVAns1Xu22RRfv"}]}]}} {"custom_id": "Rajbari_17955_7182_18468_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gs5nc1hqGy9P6b4DUyLmm3"}]}]}} {"custom_id": "Naogaon_16416_15903_16929_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NAU712StkmaSWgwLfGUZvg"}]}]}} {"custom_id": "Mymensingh_21546_1026_22059_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TxV5REXDU7KRkpG2eNS2Q4"}]}]}} {"custom_id": "Natore_16929_20007_17442_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EBXCPEcJnMi24tioMZQn1R"}]}]}} {"custom_id": "Habiganj_12312_13851_12825_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MgGbRAT1p6XX6GFB5hgxLG"}]}]}} {"custom_id": "Kushtia_15903_7182_16416_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U32SL8uqhvRV9Fhyjym864"}]}]}} {"custom_id": "Bogra_26163_21546_26676_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MWik4MBF23M2YbPknJDghs"}]}]}} {"custom_id": "Chittagong_27702_20520_28215_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4CbKwk7mA8sgjMWtdP9Mq3"}]}]}} {"custom_id": "Narayanganj_3591_10260_4104_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SagKvxjzt99QDZF2TPSt5F"}]}]}} {"custom_id": "Sylhet_32832_12312_33345_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M3wdQMRM9PgG2kWoRAANdu"}]}]}} {"custom_id": "Munshiganj_13851_6156_14364_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J1ccc84jhCGgkaq7akNKG9"}]}]}} {"custom_id": "Gopalganj_6669_6156_7182_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Eh2ScVEgXKsDLrec1F8Ve3"}]}]}} {"custom_id": "Jessore_4617_12312_5130_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D9aPC8zYxiDnEp2bZX98oL"}]}]}} {"custom_id": "Noakhali_14364_29241_14877_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BfYKk3KxF8faHPMJnCh8SW"}]}]}} {"custom_id": "Netrakona_16929_4617_17442_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XwnGQfntSp5x3JKtZzjar5"}]}]}} {"custom_id": "Lalmonirhat_10773_7695_11286_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HzuTxvuQwc4FPGM6ugXEMj"}]}]}} {"custom_id": "Comilla_19494_21546_20007_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GBF4UiCG1YJrzr5SPfK5S7"}]}]}} {"custom_id": "Manikganj_17442_10773_17955_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TDvScrP96LwWq3Pr9WL1TC"}]}]}} {"custom_id": "Bandarban_2052_30780_2565_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Aq6tpt1pgjqAb3QNFBHjix"}]}]}} {"custom_id": "Mymensingh_20520_28215_21033_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TaAu8Bz7NsEJJpKuopaaCe"}]}]}} {"custom_id": "Patuakhali_18468_38988_18981_39501.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AYpbimuKMb8BTvePXps5sg"}]}]}} {"custom_id": "Natore_18981_10260_19494_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SEdJD2u6ciXbk3CSSNvBb2"}]}]}} {"custom_id": "Sunamganj_11799_18468_12312_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1VYf4MM8MCbRksYozvfqnW"}]}]}} {"custom_id": "Manikganj_12825_5130_13338_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RZeXqmtMcc9XwquNHVoFii"}]}]}} {"custom_id": "Patuakhali_18981_17955_19494_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ewa2py4CJYAWogHiVG6bzu"}]}]}} {"custom_id": "Patuakhali_20007_30267_20520_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RZfZqnKskjLVDu6PKtfaGQ"}]}]}} {"custom_id": "Sirajganj_22572_33345_23085_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Wceazf4eTXAbCt83EbxRuo"}]}]}} {"custom_id": "Sirajganj_23085_34884_23598_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2FuVk7Lyr8iebQjqe9Qztu"}]}]}} {"custom_id": "Lakshmipur_5130_16416_5643_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Jiq92VV9Z8X1PkD71Kb3d4"}]}]}} {"custom_id": "Sirajganj_23598_9234_24111_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RHk1yoZ2FnKDAADCz8z9Qp"}]}]}} {"custom_id": "Dinajpur_17955_12825_18468_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7SkADnNbU17BvVmDqjHETd"}]}]}} {"custom_id": "Naogaon_17442_6156_17955_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5BYuK56dt3f36W2uXpVrtH"}]}]}} {"custom_id": "Sherpur_5643_1026_6156_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V42hQc6GpGa9xXxTQN8Tbg"}]}]}} {"custom_id": "Gaibandha_20520_29754_21033_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FqTeGWkayPiCHY9pv7SYXE"}]}]}} {"custom_id": "Rangamati_513_53352_1026_53865.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GmiTXTWhvss9ZhPM3GvYPS"}]}]}} {"custom_id": "Rangpur_14364_10260_14877_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KenEi2JVpjJMw87WJTkNwE"}]}]}} {"custom_id": "Gopalganj_16416_5130_16929_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7W498GPcZYzTGzsfFwUDw1"}]}]}} {"custom_id": "Chuadanga_5130_12825_5643_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L2J7ZZPkW7ZtLEwtJm46Ee"}]}]}} {"custom_id": "Mymensingh_8721_12825_9234_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PT7osn3vwtJrfKwHVfq8sS"}]}]}} {"custom_id": "Jhenaidah_23598_10773_24111_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4zgNGbR39emTBWP8UHmeeK"}]}]}} {"custom_id": "Rajshahi_15903_15903_16416_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4jRGF6rhJSJzrvjT7LWKin"}]}]}} {"custom_id": "Bandarban_23598_40014_24111_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QS1YmcniaMRtjc2USE8yR3"}]}]}} {"custom_id": "Jhenaidah_14877_20520_15390_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NiKDgT7mWqZi8BaUA5YY2D"}]}]}} {"custom_id": "Satkhira_9747_53865_10260_54378.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P6rkgQ9vPbgz3obm7WYugS"}]}]}} {"custom_id": "Khagrachhari_15390_9234_15903_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HCaEYKLuV3wVr36WDLEPob"}]}]}} {"custom_id": "CoxsBazar_6669_5643_7182_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BrPZNtpYURd9EExFS2hqPd"}]}]}} {"custom_id": "Khagrachhari_16929_13851_17442_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QDgn2xrCgxnhWkkANmrPhL"}]}]}} {"custom_id": "Jhenaidah_24624_22572_25137_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5crjeLj1nM9ZHPgXeup6oA"}]}]}} {"custom_id": "Rangpur_6156_9747_6669_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BHQSANJ6vCs7TNjC4F5r33"}]}]}} {"custom_id": "Jamalpur_8208_34884_8721_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L4YK3sRbHs8pcGarTqwa7e"}]}]}} {"custom_id": "Jhalokati_5643_4104_6156_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T1PFYVji2cLYuLhh66WZNW"}]}]}} {"custom_id": "Sylhet_35910_16929_36423_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VFi7KEfC6XWTpQcYfpAHry"}]}]}} {"custom_id": "Pirojpur_10260_9234_10773_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-67epKDYcZVebLjee7rH9qq"}]}]}} {"custom_id": "Joypurhat_12312_4617_12825_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Wm4qnnbzbqCWuFoaEXot7Q"}]}]}} {"custom_id": "Bogra_28215_6669_28728_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HY3bP7MR8Yhoeq6BkYCTAe"}]}]}} {"custom_id": "Khulna_2052_24111_2565_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SRL3J5hjWe6NkeKWwtSNtw"}]}]}} {"custom_id": "Naogaon_10260_17442_10773_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CfN2W96aN9S5hebLizQVNa"}]}]}} {"custom_id": "Rangamati_10773_9747_11286_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-34DNr1yshBs2NQWu9Zk5JW"}]}]}} {"custom_id": "Bandarban_21033_45657_21546_46170.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-272Eg7iX3c2sjJYGEN65hs"}]}]}} {"custom_id": "Pabna_11286_5643_11799_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6zBzaWPf3MKBVbfhiRkfXZ"}]}]}} {"custom_id": "Meherpur_6669_5130_7182_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VYVdTAbwgV2HcTFb4q7euq"}]}]}} {"custom_id": "Rajbari_3078_7182_3591_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LyHcfpdKoYQ8GjL8kQkLVH"}]}]}} {"custom_id": "Sunamganj_21033_8721_21546_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XLwNTzxezRwEM4G7AR2Mbz"}]}]}} {"custom_id": "Narsingdi_7182_19494_7695_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5w77Mmu2AJHwXzgbGeG3qY"}]}]}} {"custom_id": "Narail_6669_13338_7182_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JKUXbFxjMPefDQrPZddA4e"}]}]}} {"custom_id": "Brahamanbaria_11799_18981_12312_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TdpwkYfGjiHVNTqr33CryA"}]}]}} {"custom_id": "Patuakhali_24624_15390_25137_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CXgih7dnKLt5F95qeWyfAr"}]}]}} {"custom_id": "Tangail_9234_22059_9747_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KpjxgsFS3DaPPkLeambhe5"}]}]}} {"custom_id": "Bandarban_12825_44631_13338_45144.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LRMu5MLue9rLytKaQZtd1F"}]}]}} {"custom_id": "Khulna_8721_59508_9234_60021.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T5CUPx5Yjq4SnRZmAKoqUy"}]}]}} {"custom_id": "Chittagong_2052_16416_2565_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5HzWCVzGjtFocnqJodHVKu"}]}]}} {"custom_id": "Bagerhat_10260_24111_10773_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EMSGLS9X19MoVnkVE5s41e"}]}]}} {"custom_id": "Pabna_28728_26676_29241_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MuLEVzDitChgmp2UJCNZVa"}]}]}} {"custom_id": "Nilphamari_13851_7695_14364_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R34MBzCQ4fpaxVxGnx2tuK"}]}]}} {"custom_id": "Madaripur_11799_513_12312_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2SWmnS1s4Q93j46jCxtsCq"}]}]}} {"custom_id": "Gaibandha_11799_29754_12312_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TudztPxgE1o1n4WRLe79N1"}]}]}} {"custom_id": "Rangpur_11799_18981_12312_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HWVSg6sAQPfnYY9VEnmHwf"}]}]}} {"custom_id": "Sunamganj_32832_10260_33345_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5J4MwPPHkDfskP3s4J23sP"}]}]}} {"custom_id": "Pabna_22059_18468_22572_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Q2SKLki5UK9pgp1LEAdkHg"}]}]}} {"custom_id": "Netrakona_28215_22572_28728_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AJjp8hvYDo4ATEjADCP3k3"}]}]}} {"custom_id": "Bagerhat_3078_58995_3591_59508.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9KQDueE2zuzeJR7xE9oVXv"}]}]}} {"custom_id": "Barguna_11799_12312_12312_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KRqKxLK9ew9JxZgQKfdBCQ"}]}]}} {"custom_id": "Madaripur_9747_6669_10260_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SyRj5wEft9y9G4fKWHRs5p"}]}]}} {"custom_id": "Gaibandha_21546_9747_22059_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3LyYtsn1NXg64bG1gZoqFt"}]}]}} {"custom_id": "Bhola_4617_11799_5130_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6uvJHsWwCE35oQua6AGRkP"}]}]}} {"custom_id": "Habiganj_17442_9234_17955_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8zxMMSYaj4Na2fJrV9o42x"}]}]}} {"custom_id": "Tangail_4104_33858_4617_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SqtyKCFVKKVgurSqg5Cjbh"}]}]}} {"custom_id": "Khulna_4104_35397_4617_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FqW1kqtXmbc2eXv5JdtR9F"}]}]}} {"custom_id": "Rajbari_20520_9234_21033_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XAMN4FeDjUMLXTYrjojUpd"}]}]}} {"custom_id": "Brahamanbaria_17442_9747_17955_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2vYic6SdA39Z8BBp7GtZNM"}]}]}} {"custom_id": "Comilla_21546_7695_22059_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-34EBXJCpiqfi5p51VuSXSz"}]}]}} {"custom_id": "Barisal_12825_10260_13338_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3SSTPEVGP6UFukdKTHoaPj"}]}]}} {"custom_id": "Netrakona_8208_3591_8721_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JHzGGyv6pmrQ4AAhuh24pX"}]}]}} {"custom_id": "Kushtia_18981_19494_19494_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BHR7YrnKJ3vcBfD9tBnYgF"}]}]}} {"custom_id": "Mymensingh_3591_25137_4104_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8XdCJ48Eiy16TBY8aArNVX"}]}]}} {"custom_id": "Bagerhat_7695_28215_8208_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UdGPohG1TPzji5Ka7VSpMx"}]}]}} {"custom_id": "Barisal_12312_15390_12825_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CE1oL8tfZctHRpkdUb1Gpx"}]}]}} {"custom_id": "Narail_9234_4104_9747_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7LGueCAu9CWsEW55btGpr9"}]}]}} {"custom_id": "Mymensingh_2565_27189_3078_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PY31D29hpe5z2VUKPzMnYd"}]}]}} {"custom_id": "Manikganj_25137_13851_25650_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FKwnJQ9P1dwWPbeBv32AeP"}]}]}} {"custom_id": "Habiganj_15903_24624_16416_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1rQMfJGz12WQ1uAfRkTxAX"}]}]}} {"custom_id": "Dinajpur_14877_19494_15390_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Aa9CUJ4QkB1ApJ9RHsWH8z"}]}]}} {"custom_id": "Bagerhat_12312_23085_12825_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AXyZCacCNh8cuU18fGtUkW"}]}]}} {"custom_id": "Natore_14877_2052_15390_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Jz9t7ZDR2EV5DKnFWoc4wY"}]}]}} {"custom_id": "Jamalpur_18468_30267_18981_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6EwRedqMG5pWEm7h4hDFdb"}]}]}} {"custom_id": "Barguna_4617_24624_5130_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3WWCA1coZWhGK2dUumg74j"}]}]}} {"custom_id": "Kurigram_11799_42579_12312_43092.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1hckFKj2WSm3ZpNEnv5oxj"}]}]}} {"custom_id": "Comilla_25137_17955_25650_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BNQ9g6YpzVQDSCTMw7M2Vc"}]}]}} {"custom_id": "Narayanganj_4617_12825_5130_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bag9k3tEMynxhFa9q5hyQq"}]}]}} {"custom_id": "Rajshahi_21033_6156_21546_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9NuqeQqE15kanFWjaqMZiU"}]}]}} {"custom_id": "Jhenaidah_28215_4104_28728_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RkUmuF8AvJndtj5m2FiL5E"}]}]}} {"custom_id": "Gazipur_3078_12825_3591_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GNbHXmfZ7KLfQmjwcfiFNA"}]}]}} {"custom_id": "Patuakhali_5643_18468_6156_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mc6o3UNZer27MLXKn4K2b6"}]}]}} {"custom_id": "Panchagarh_15903_30780_16416_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4BJxJ7S8zzueaeddZD5yq5"}]}]}} {"custom_id": "Rangpur_16416_20007_16929_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8hpZTbhm91wsSxdwu91YEZ"}]}]}} {"custom_id": "Tangail_16929_37449_17442_37962.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Xmo85pRKkp4cVQxRCQ8mFF"}]}]}} {"custom_id": "Mymensingh_17442_6669_17955_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2Ckd6UW6TAQmRcCFhBwM41"}]}]}} {"custom_id": "Noakhali_6669_9747_7182_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6gyF9sE3F51Qpd6YLmcnXZ"}]}]}} {"custom_id": "Pabna_16929_19494_17442_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4TRuoAsNVu1is2YWMVvJrT"}]}]}} {"custom_id": "Narayanganj_7182_10773_7695_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KXCGdAyh5oMr9UEkJyJyQ2"}]}]}} {"custom_id": "Mymensingh_2052_21546_2565_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QkvXDtfhmEq1xiuq63yMcy"}]}]}} {"custom_id": "Rangamati_15903_52839_16416_53352.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DqUBrLphYhM1sr7F4zhjWn"}]}]}} {"custom_id": "Lakshmipur_11286_21033_11799_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vffdk5pBGauVaUwZdkCX3C"}]}]}} {"custom_id": "Joypurhat_4617_5643_5130_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8aHfSHQK9L4kzK2gXzDsVB"}]}]}} {"custom_id": "Comilla_5130_10773_5643_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V7MAg3CmnQAN12TrF8jLKW"}]}]}} {"custom_id": "Gazipur_6156_16929_6669_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4nqhGoJEMSdvRX8phQ1g1y"}]}]}} {"custom_id": "Rajbari_16416_13338_16929_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QTntNpdoCUbNVnaQi6FBQY"}]}]}} {"custom_id": "Habiganj_13338_21033_13851_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XGeDi5Gvzs5N1dggNaysdA"}]}]}} {"custom_id": "Rangamati_17955_62073_18468_62586.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8ZQTL1YJSwBzpC4ziN33ns"}]}]}} {"custom_id": "Bogra_11799_17442_12312_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2S4KTwkZ7b9Ckd4ittQY11"}]}]}} {"custom_id": "Lakshmipur_5643_17955_6156_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HAjXzpuryiZiLMAgXEo2JN"}]}]}} {"custom_id": "Bogra_14364_13338_14877_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NbvzYTaoZN6Ty7sgQCHgwx"}]}]}} {"custom_id": "Sirajganj_12312_21033_12825_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HjfB7NpaYcGCc3TN9DrwrL"}]}]}} {"custom_id": "Madaripur_8208_15903_8721_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RajxzUwygVTVL1VJr626we"}]}]}} {"custom_id": "Rangamati_16929_66690_17442_67203.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-E8cEhUhoV51165X7eYBP7E"}]}]}} {"custom_id": "Naogaon_15390_513_15903_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PnCJeEsmd3EDuSa4Fri5oF"}]}]}} {"custom_id": "Bandarban_10260_20007_10773_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CDeewFba2TYNg3qnEs1vYH"}]}]}} {"custom_id": "Rangamati_7695_47196_8208_47709.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TRQFg5SqKVF5L7RwEFVi6f"}]}]}} {"custom_id": "Chittagong_10773_4617_11286_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2dZBWcpr9ScHPHMvMDdD3v"}]}]}} {"custom_id": "Jamalpur_16929_28728_17442_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VuQUm6FXyUgggnNepJtr6u"}]}]}} {"custom_id": "Satkhira_10260_60021_10773_60534.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GJkyPW8eUkfFsnZJnY1Qzo"}]}]}} {"custom_id": "Sherpur_11799_14877_12312_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LhPNLsvnxUYaki5oWL443A"}]}]}} {"custom_id": "Chittagong_21546_16929_22059_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NGGQBNFKi99sJnKcdD4T8K"}]}]}} {"custom_id": "Satkhira_15390_41553_15903_42066.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HEr3PQ2U4hqQNEt3HuBPCA"}]}]}} {"custom_id": "Sunamganj_23085_21033_23598_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NCBaUuzoqmihCoFjFLfEt5"}]}]}} {"custom_id": "Natore_10260_13338_10773_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SCEyL8vieAFeHVwZEGhNaB"}]}]}} {"custom_id": "Patuakhali_20520_30780_21033_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BiTPTcb1kmUWx84aZeCGF4"}]}]}} {"custom_id": "Noakhali_7182_8721_7695_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-19BjWgAN8mt7hYVoDEG4gB"}]}]}} {"custom_id": "Dinajpur_13338_11799_13851_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UenEQJHvopGkaof8MYWdAk"}]}]}} {"custom_id": "Maulvibazar_4104_32832_4617_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6hgBycnQ6RtH6UZ2ZYPfvi"}]}]}} {"custom_id": "Jhenaidah_22059_11286_22572_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Frax8jdGjt57FTdRG6D68a"}]}]}} {"custom_id": "Pabna_31806_27189_32319_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ANXKzFxkpZEVWh61vJecY6"}]}]}} {"custom_id": "Mymensingh_20520_20520_21033_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FhbpqJ1tQ7vYh1By1ogB8t"}]}]}} {"custom_id": "Narayanganj_7695_9747_8208_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7yJd5tkjWDo3EykL3XY9p4"}]}]}} {"custom_id": "Naogaon_23598_1539_24111_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KB28nsDzUk8p9a2M8vKC3c"}]}]}} {"custom_id": "Madaripur_9234_18981_9747_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6Pdxd9nsKCWQ1gZcz4Ygh4"}]}]}} {"custom_id": "Kushtia_15903_7695_16416_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Tj1DSCjRCPMEkvyPczhum6"}]}]}} {"custom_id": "Satkhira_13851_46683_14364_47196.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XdRxEueYuynTQwCGPtnWRM"}]}]}} {"custom_id": "Pabna_13851_5130_14364_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5uLq5xoNGxkmWRYbrYJUze"}]}]}} {"custom_id": "Noakhali_12312_34371_12825_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1wpoAx38SZC5qNxVBZvWPN"}]}]}} {"custom_id": "Sherpur_13338_18468_13851_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LjCKfMJYMC3JHAjTk6V17w"}]}]}} {"custom_id": "Natore_15903_9747_16416_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9gqneTK3Uba7pXm2oG1KHn"}]}]}} {"custom_id": "Rajshahi_8721_18468_9234_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SMU1phDiUmeWgKrMYe9obD"}]}]}} {"custom_id": "Bhola_6669_41040_7182_41553.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PWQdXpeGQLRh21PDqG5fA7"}]}]}} {"custom_id": "Dinajpur_33345_36423_33858_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AZwkKrfSyRNWG3H2D62jt9"}]}]}} {"custom_id": "Faridpur_21546_8721_22059_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ad9dxEsfeMcpDENjVfNdhw"}]}]}} {"custom_id": "Mymensingh_14364_30267_14877_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5g5bNccW51Hpxqqub3NxH9"}]}]}} {"custom_id": "Kushtia_21546_13338_22059_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BSzpCLDCMRPaEfd4YRVFRU"}]}]}} {"custom_id": "Joypurhat_1026_3078_1539_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HupY5v8Y7Uv4j4TVfkZbXL"}]}]}} {"custom_id": "Gaibandha_8721_20520_9234_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WwWLaqYskt5nugb4unEkWh"}]}]}} {"custom_id": "Kishoreganj_15903_9747_16416_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XRrrEtkRvSvxwD9vXmgFFq"}]}]}} {"custom_id": "Nawabganj_1026_14877_1539_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GdoW2z6mWdwheuS1Ddpocs"}]}]}} {"custom_id": "Habiganj_9234_35397_9747_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D7mPmsXRFMyfrPCfRih9i2"}]}]}} {"custom_id": "Gaibandha_513_24624_1026_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7EBQF3aymWzYLjL3xJqNoD"}]}]}} {"custom_id": "Bagerhat_513_17955_1026_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SJywPBYHjQj7uqgjR9XfwR"}]}]}} {"custom_id": "Meherpur_9234_13851_9747_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Qiy3ibnqp77BdVjDaqVYDR"}]}]}} {"custom_id": "Bogra_18981_8208_19494_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UoFGkfuAaw2iJaYr7cTQbq"}]}]}} {"custom_id": "Gaibandha_19494_15390_20007_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6TVdxNVN9eABoNKTwM74FQ"}]}]}} {"custom_id": "Rajshahi_18468_18468_18981_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AhLzaZmAzTSjjoWbNNb9P2"}]}]}} {"custom_id": "Kurigram_13338_25137_13851_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Jm4pY2hu3N9SdC7BEXW5Fx"}]}]}} {"custom_id": "Rajbari_20520_15390_21033_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GdNUt2dcer7Cfg4U5cCWjA"}]}]}} {"custom_id": "Rangamati_30267_80028_30780_80541.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GUjiTQPgGCqsWiafv8Q5vm"}]}]}} {"custom_id": "Gaibandha_9234_10260_9747_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3vKnmV1Ljn26CT5Z7aSGDn"}]}]}} {"custom_id": "Bhola_9747_30780_10260_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D55sHUUL63kjTm5D6zyYLB"}]}]}} {"custom_id": "Shariatpur_17442_12312_17955_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5qdjgmNAeUfL5LCn7LQr5D"}]}]}} {"custom_id": "Magura_7695_15390_8208_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RMcTBTcgVv74KWun2fVxeG"}]}]}} {"custom_id": "Dinajpur_7182_5643_7695_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Wvw6kxgddPJCjewYmNomp6"}]}]}} {"custom_id": "Narail_7182_3078_7695_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2qQSyEqYq3rHrFZ2oGa7r3"}]}]}} {"custom_id": "Nilphamari_4104_3078_4617_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-C6vAarmskJG1eUWZfy27Td"}]}]}} {"custom_id": "Maulvibazar_18468_9747_18981_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W56PVXjwkdwr5vvLK646VB"}]}]}} {"custom_id": "Gazipur_8721_19494_9234_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2Evc9rd2bwYY2GXQr6iHb9"}]}]}} {"custom_id": "Sylhet_13338_7182_13851_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4fZZ62cjmvRHFJ3S5YKZxB"}]}]}} {"custom_id": "CoxsBazar_15903_40014_16416_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RtwowT17t6xDSHhLF1c7ru"}]}]}} {"custom_id": "Nawabganj_13851_13851_14364_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QqvLruaZ4d3oAjCo5t7qfz"}]}]}} {"custom_id": "Kishoreganj_22059_4617_22572_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MK9Vb4o4iw7xwSJFAo5jqx"}]}]}} {"custom_id": "Gaibandha_17442_27189_17955_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RpxEMn8m8C1gisW77kNtPJ"}]}]}} {"custom_id": "Panchagarh_16416_15903_16929_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TEMGTu5RyneoqvMWafWYpt"}]}]}} {"custom_id": "Meherpur_2052_5643_2565_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4hzY1M1tzuM77bveNyXa2r"}]}]}} {"custom_id": "Thakurgaon_513_18468_1026_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QpM3feHT8EaZZDHGZza1yQ"}]}]}} {"custom_id": "Khagrachhari_5130_19494_5643_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-H9eMres6BQdgYAK1kZQVY7"}]}]}} {"custom_id": "Madaripur_10773_5130_11286_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Sp3BkLZibXb232EcYaq89k"}]}]}} {"custom_id": "Habiganj_15903_19494_16416_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3xLtcEsTJe4nyLfM2R1Njs"}]}]}} {"custom_id": "Lakshmipur_6669_12825_7182_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CZN5A1k7iQMfrGBFbs4mr4"}]}]}} {"custom_id": "Noakhali_5643_19494_6156_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GLDeqmHuvDrB76puJujoam"}]}]}} {"custom_id": "Tangail_11286_39501_11799_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CCwKr9NCGTtAeRrRmErBqp"}]}]}} {"custom_id": "Rangpur_9747_28728_10260_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SG3tdzaQVtcU2EiboYCEuo"}]}]}} {"custom_id": "Rangpur_21546_12312_22059_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KXsc6jfZ9CR6fpEvKQPHqh"}]}]}} {"custom_id": "Nilphamari_6669_25650_7182_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UVbruVxvNQx1JGwBoqS47i"}]}]}} {"custom_id": "Munshiganj_11286_4617_11799_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CimLKYjcboeie4bgSr6mDB"}]}]}} {"custom_id": "Thakurgaon_18981_2052_19494_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MYDuzDJ5Gzsqv98TrNnLYn"}]}]}} {"custom_id": "Satkhira_2052_19494_2565_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9PaeAu2WxrHVSNBRUbZL9Q"}]}]}} {"custom_id": "Rangamati_14877_17955_15390_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Pvu73jyL4HHauZswGnWKBR"}]}]}} {"custom_id": "Gaibandha_25650_17442_26163_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9fEdV9TYefmFSZmQjhaYbN"}]}]}} {"custom_id": "Bogra_13338_17442_13851_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RR9K21K78cyDLonQtzNsgD"}]}]}} {"custom_id": "Satkhira_14877_6156_15390_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LjpZQFEKQkmWy5ufjE9KA9"}]}]}} {"custom_id": "Sunamganj_34371_4104_34884_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MRkdjBiu8ksBNBLrAiD8Np"}]}]}} {"custom_id": "Jessore_16929_13851_17442_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XAXkbJ2wvBFG2oppr8gg3t"}]}]}} {"custom_id": "Sunamganj_18468_2052_18981_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PpzXXYKxKtXvaUDeUChVsq"}]}]}} {"custom_id": "Jhalokati_11286_10260_11799_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QbMBAGfNN2W6y2YrtLPTPj"}]}]}} {"custom_id": "Munshiganj_18981_10773_19494_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AFwStuyV3yFzqv779ASwyo"}]}]}} {"custom_id": "Naogaon_24111_20007_24624_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Lb6FmXVfcgVpm3tHR4Kmq7"}]}]}} {"custom_id": "Kishoreganj_19494_27189_20007_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L8i4Cf1YqdCFdnBYBzg9Jq"}]}]}} {"custom_id": "Khulna_8721_63099_9234_63612.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D6MyM2oLdbvqzcHE6EdCBc"}]}]}} {"custom_id": "Chittagong_33345_51300_33858_51813.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VTbPE5w3vrfJJxvWYRNQkd"}]}]}} {"custom_id": "Chandpur_5130_14877_5643_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6ktQ31bNHWL1b8thWZjH4R"}]}]}} {"custom_id": "Tangail_13338_30267_13851_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HEox9SJQjHPyQiGedam1pm"}]}]}} {"custom_id": "Tangail_10260_11286_10773_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2xMymgPwqHMcDUw53KCQA9"}]}]}} {"custom_id": "Kushtia_29754_16416_30267_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P3jYksbZb4opunh3dm6CWF"}]}]}} {"custom_id": "Patuakhali_23085_11799_23598_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6dvm2kv4SYpsPBnJQ3c3sd"}]}]}} {"custom_id": "Magura_5130_13338_5643_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AWsgyzo1rQ1BP6M2zqsWaN"}]}]}} {"custom_id": "Bandarban_9747_41553_10260_42066.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ASrxor6RyZgqu2yhGkeQGp"}]}]}} {"custom_id": "Chittagong_19494_8721_20007_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9UtpHbUWn6WjiMpnJLFSQr"}]}]}} {"custom_id": "Gopalganj_10773_12825_11286_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GL6EW3BhqwRdNE8W5oiNws"}]}]}} {"custom_id": "Tangail_12825_32319_13338_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BUKkG1NBxYWdZpEBHpMDqm"}]}]}} {"custom_id": "Madaripur_14364_20520_14877_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-91UwHhhuzHuxfngU95NKAT"}]}]}} {"custom_id": "Khulna_1026_15390_1539_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QxVE2kwDkZcGzjotnCCSHs"}]}]}} {"custom_id": "Lakshmipur_11286_14877_11799_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AHrZccjX2rQ5rS8JS3D8KV"}]}]}} {"custom_id": "Bhola_2052_13851_2565_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MJjkshXnP7NSbBoXimMUJB"}]}]}} {"custom_id": "Rangamati_7182_3078_7695_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QgNbSB3REzfC159X8wM9Vf"}]}]}} {"custom_id": "Pabna_23085_19494_23598_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GxznNTwNw57ZJDxxSGzAZs"}]}]}} {"custom_id": "Rangpur_13338_11286_13851_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XX2CuUPUbHNUD4Bey5xpzF"}]}]}} {"custom_id": "Bhola_1539_4617_2052_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R88Y8RFqA3zvRgQ1CVJv5y"}]}]}} {"custom_id": "Jhalokati_3591_21546_4104_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SEdQUkdjqALfcSQxSpXpJh"}]}]}} {"custom_id": "Mymensingh_20520_46683_21033_47196.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SRQMbWdZcTVTrjzkjjB6Qf"}]}]}} {"custom_id": "Kushtia_12825_19494_13338_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Qi6K2mYHfZBZN5wqpbaTq2"}]}]}} {"custom_id": "Gopalganj_11286_22059_11799_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WDriyHbkcXZ2G5a6mAV84S"}]}]}} {"custom_id": "Brahamanbaria_21033_14877_21546_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ppdzs7wEyZHH8Sm9bJGdca"}]}]}} {"custom_id": "Naogaon_25650_25650_26163_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JiELBSEUZMY4eHAps2uToS"}]}]}} {"custom_id": "Habiganj_20520_21033_21033_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ufz8xpzW2ucnGUueqwdi1j"}]}]}} {"custom_id": "Barisal_24624_2052_25137_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Y2TbgTUGFcPqLun1u9S746"}]}]}} {"custom_id": "Dinajpur_26163_33858_26676_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NSMpjbUUPkVYFUZqfEom93"}]}]}} {"custom_id": "Satkhira_17442_12312_17955_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PWL54PXciAPb6jx97dqWAS"}]}]}} {"custom_id": "Manikganj_10773_17442_11286_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BMcgKFBk2h2wWsn3hxAX62"}]}]}} {"custom_id": "Satkhira_17955_52839_18468_53352.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WFYHWNGJYNAMguWh59drGt"}]}]}} {"custom_id": "Bandarban_13338_34371_13851_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XZ6RtiQ9FnUYUEfdrXUwKX"}]}]}} {"custom_id": "Chittagong_25650_35910_26163_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KPUwao49wK8FXHZjRxboe4"}]}]}} {"custom_id": "Rajbari_3078_4617_3591_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Cgak7JH33fArsxZNJmDNkj"}]}]}} {"custom_id": "Magura_8721_6669_9234_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PwPCrv9RTZY6J4d3LGMsPc"}]}]}} {"custom_id": "Pabna_10260_18981_10773_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LxWDXyrDmsgboyvQ7kXiUa"}]}]}} {"custom_id": "Lakshmipur_8721_3591_9234_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CgSCewzc7fkwHUMBrfkX3B"}]}]}} {"custom_id": "Brahamanbaria_20007_20520_20520_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A8pys2LvpBmrtDDWF1idAe"}]}]}} {"custom_id": "Mymensingh_18981_40014_19494_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X2661aevcVQ5hTgQFgpB4L"}]}]}} {"custom_id": "Jessore_16929_3078_17442_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5UvejzraMATSXucN5VUQme"}]}]}} {"custom_id": "Munshiganj_16929_13851_17442_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QwqfPQKMcLgortRPydapYB"}]}]}} {"custom_id": "Lakshmipur_1539_10260_2052_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MYUNxiHDCDNk5QwkWdUB2d"}]}]}} {"custom_id": "Bogra_10260_19494_10773_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Cy9fXwsaDXmsZ1p5T8mNFX"}]}]}} {"custom_id": "Bandarban_9747_27189_10260_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CAitmXjYfzbMP3FBiKcR7p"}]}]}} {"custom_id": "Comilla_31293_28728_31806_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CZMDibACm9L3wdrx1RpcWv"}]}]}} {"custom_id": "Natore_2052_8208_2565_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-99gsRm1F2WVCQexZ5617RP"}]}]}} {"custom_id": "Feni_7182_7182_7695_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QyaxBGTShVjVQRsyCGNVB2"}]}]}} {"custom_id": "Habiganj_16929_15390_17442_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UrUSrFMEvd5gvzXyxkyACN"}]}]}} {"custom_id": "Khulna_11286_7695_11799_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9TbzT5TuvcSXcXdUpVNhXt"}]}]}} {"custom_id": "Chittagong_9234_26676_9747_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FYyTCvkKdwZSq7ks5EjPjZ"}]}]}} {"custom_id": "Maulvibazar_8721_29754_9234_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-589Rf5w9H2rsky1BmSWEU1"}]}]}} {"custom_id": "Gopalganj_16416_22059_16929_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5RTU8zjxdwAZ7g9Kzyzt9n"}]}]}} {"custom_id": "Rangpur_1026_11286_1539_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6Q6yEsT8rAPR4PWTjSa8yz"}]}]}} {"custom_id": "Comilla_20007_10260_20520_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SzAod65stL99ueaGRUuLrP"}]}]}} {"custom_id": "Dinajpur_9234_13338_9747_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TBGbEyDecmbuuJR5nVeLNT"}]}]}} {"custom_id": "Narail_6156_12312_6669_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KV2QJ5TmDCv1628JtdcPKm"}]}]}} {"custom_id": "Gaibandha_20520_21033_21033_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UhhUjWSbzFfKVpuwNrH5sv"}]}]}} {"custom_id": "Nilphamari_15390_18468_15903_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mqh33wnjQyjmUpuLSdCbrc"}]}]}} {"custom_id": "Jhenaidah_26163_15903_26676_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1t4bo4DJxgHVX1oSWhk9TQ"}]}]}} {"custom_id": "Rajshahi_17442_20007_17955_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LdnBjTqREFTpGr9iugHcdj"}]}]}} {"custom_id": "Mymensingh_9234_30267_9747_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gt7kZvi6yHv6sktCRynu9C"}]}]}} {"custom_id": "Mymensingh_22059_27702_22572_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NFMbzwYvthJ7aj2Sryczs6"}]}]}} {"custom_id": "Narayanganj_6156_16929_6669_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GVrcVj5uE4z6AGN3B85Lkz"}]}]}} {"custom_id": "Netrakona_27702_18468_28215_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WFt4hLh49LqSR8faDBFt1B"}]}]}} {"custom_id": "Bagerhat_10773_17442_11286_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Jvfuaj1Qw2DyALxjADUvwU"}]}]}} {"custom_id": "Panchagarh_16929_14877_17442_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BUQ2J6Y65mZBkjuhc3HwXt"}]}]}} {"custom_id": "Comilla_27189_20007_27702_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NdaxQdpJ8eakqqyKzSPmRu"}]}]}} {"custom_id": "Panchagarh_15390_14364_15903_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6dtbx86r7N7tXynFR67WEc"}]}]}} {"custom_id": "Nawabganj_5643_14364_6156_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FP7da4dTiri7roWLtYbHyB"}]}]}} {"custom_id": "Jamalpur_1539_25137_2052_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bhhy8CgYhY9AjEYxBZmPxi"}]}]}} {"custom_id": "Tangail_14877_29241_15390_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DtDojVQnw1cfCm8jsMXCcp"}]}]}} {"custom_id": "Chuadanga_2565_10773_3078_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JKML9prFUuXjf5a4EyjrxN"}]}]}} {"custom_id": "Joypurhat_4104_10773_4617_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XUxp5gzuHd2WXUM2gLZcxT"}]}]}} {"custom_id": "Sunamganj_9747_6669_10260_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JNSEPTuSsyt9dirAcoiSEG"}]}]}} {"custom_id": "Rangpur_15903_7695_16416_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CTW4XRZghQfFJX8KZN9PcN"}]}]}} {"custom_id": "Bhola_18468_37449_18981_37962.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WJoFddAzsR36bgkTRfzoL4"}]}]}} {"custom_id": "Natore_7182_18981_7695_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9izarNygtemPJPsuJeV7sP"}]}]}} {"custom_id": "Shariatpur_0_8721_513_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RR589Nc2M7KW6hz7smwvoz"}]}]}} {"custom_id": "Bogra_6669_18468_7182_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vdjhx2RPFNdxJzERETGPwD"}]}]}} {"custom_id": "Jhenaidah_9234_24624_9747_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NWyjguxunGDUUC9MaMdTUs"}]}]}} {"custom_id": "Gopalganj_9234_3078_9747_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RoAzyVxS3YDDdjR8R3y6r8"}]}]}} {"custom_id": "Lakshmipur_8208_12312_8721_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QsLPMb3Sj3MVikKYQ32G5M"}]}]}} {"custom_id": "Rajshahi_21546_17955_22059_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X1UeRB7qCB9f8tRrZZmxMk"}]}]}} {"custom_id": "Rangamati_14877_18468_15390_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W4k2eSupkYtLjUU8Uq62F9"}]}]}} {"custom_id": "Lalmonirhat_22059_27189_22572_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5aU3UFYHYLhKWPkgh8sjVp"}]}]}} {"custom_id": "Maulvibazar_15903_23598_16416_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4ALCbWrpSZU8pj8hdmKtkD"}]}]}} {"custom_id": "Barisal_4104_14364_4617_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JDNdFevwdnVKVAQUoQEXV5"}]}]}} {"custom_id": "Habiganj_6156_29241_6669_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5c1eAPYotTF6VdVAUbB8x3"}]}]}} {"custom_id": "Sherpur_1539_12312_2052_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7z3vTurKiubxoS3Es3Yb94"}]}]}} {"custom_id": "Tangail_3591_17442_4104_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-C89m7GZVXZgC8aaUXreDK9"}]}]}} {"custom_id": "Gaibandha_4104_24624_4617_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VirTPyqjhxNYxyFjEkt8y3"}]}]}} {"custom_id": "Natore_7182_22572_7695_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LHU5puwvDB5sPfu75fp7bo"}]}]}} {"custom_id": "Madaripur_12825_2565_13338_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GBMMN4wNxpPKRyfZSefVvC"}]}]}} {"custom_id": "Bogra_25650_29241_26163_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1upU8KURqzAibE6exkgMpP"}]}]}} {"custom_id": "Panchagarh_2052_17955_2565_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VsDExbKDKbQ3VQsQDPvWVK"}]}]}} {"custom_id": "Patuakhali_11286_11799_11799_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DsF4hWkm2syHXG3EVnfkjG"}]}]}} {"custom_id": "Natore_17955_15903_18468_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3HJqP4zo55FH5NMoRKdqqu"}]}]}} {"custom_id": "Bogra_19494_11286_20007_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ehpgqpg4DwaUmpSeUvm2RK"}]}]}} {"custom_id": "Dinajpur_32319_40014_32832_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DLWqAXvF2KnV7PSX1dWwPV"}]}]}} {"custom_id": "Tangail_6669_38988_7182_39501.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EM4D88Fe6rmY3Yg1XNXTSb"}]}]}} {"custom_id": "Bandarban_1539_25137_2052_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WLvXojeYnBJWpZFGSuUuvV"}]}]}} {"custom_id": "Sylhet_17442_9747_17955_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6ni81WiENh2RcenTWeMDm9"}]}]}} {"custom_id": "Manikganj_13338_18468_13851_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6DPeKpst9eFQHpB7DEDWCD"}]}]}} {"custom_id": "Rangamati_23085_67716_23598_68229.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4hFPnNAgMtBt5M1r1Px62r"}]}]}} {"custom_id": "Natore_4617_9747_5130_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HdaEZXaxrZi9ak1qjCwa8b"}]}]}} {"custom_id": "Thakurgaon_12825_22059_13338_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4rvsj79JJakEq2HsE4NzQD"}]}]}} {"custom_id": "Bogra_26676_6156_27189_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-77TmAND47HTVvsyENRgJki"}]}]}} {"custom_id": "Khulna_9747_52326_10260_52839.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9YitLBg2NecJAKijjpqGoz"}]}]}} {"custom_id": "Kishoreganj_18981_5643_19494_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-93aCqBUdbxmuZbP2wsEyuZ"}]}]}} {"custom_id": "Kishoreganj_24624_13851_25137_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BivL2BaZYNZp1U2BTBSSZH"}]}]}} {"custom_id": "Dinajpur_13338_9747_13851_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-98mt6Jb9bLXpxTGpDLBKMH"}]}]}} {"custom_id": "Noakhali_13338_11286_13851_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GK8hW7JhftQeEE9bWYPLMj"}]}]}} {"custom_id": "Pirojpur_4617_32319_5130_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3vevp7zuRM5AGzcxVAhv8H"}]}]}} {"custom_id": "Joypurhat_7695_20007_8208_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2dXnL7RKkxNPzYSAeCh345"}]}]}} {"custom_id": "Khulna_11799_54378_12312_54891.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Da1PR6tHCxVXS3bH4WqzET"}]}]}} {"custom_id": "Chittagong_31806_35397_32319_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VEusKLUCnuMVkRMNZnnN1D"}]}]}} {"custom_id": "Khulna_8721_31293_9234_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1mgddS8JsBWiKQ4iB6npfw"}]}]}} {"custom_id": "Sirajganj_13851_21546_14364_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gse1AsJjRbvtvodTN5Rp2x"}]}]}} {"custom_id": "Panchagarh_14877_26676_15390_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SgT4RUdf94LDGuCB8fCzoK"}]}]}} {"custom_id": "Pabna_32319_25137_32832_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FELkYWZVrrHAHKxwvWDLmq"}]}]}} {"custom_id": "Faridpur_17955_9234_18468_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LaEiw9135U6x84T2HHcvEk"}]}]}} {"custom_id": "Magura_14364_11799_14877_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KBjMtgi6P7vW6zWL8DrdQT"}]}]}} {"custom_id": "Natore_21033_4104_21546_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BRKCdBRZBEx9toUTit4ka4"}]}]}} {"custom_id": "Lalmonirhat_4617_6156_5130_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N2273WukDtLaig6Bym2VWT"}]}]}} {"custom_id": "Jhenaidah_18468_10260_18981_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4mvC2JPQwK34JGwJ3eKEZ4"}]}]}} {"custom_id": "Satkhira_17442_35910_17955_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CRk76ozSz8q5njL1aqM7uv"}]}]}} {"custom_id": "Mymensingh_22572_7182_23085_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GXBUi15ccr3vGA4m2nRZRN"}]}]}} {"custom_id": "Jessore_16929_15390_17442_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9kgts33JNbuVzqtjh4uDyS"}]}]}} {"custom_id": "Patuakhali_24624_8208_25137_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-93jYL3Wc8CM71nSgsw2TaE"}]}]}} {"custom_id": "Naogaon_23598_10260_24111_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W3puSQkcDxzmr8dne8EAgT"}]}]}} {"custom_id": "Sirajganj_17955_13851_18468_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BpNyJjELJfnZTgJ7eL3KNw"}]}]}} {"custom_id": "Narail_3078_8208_3591_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XjQTAWp2pFYb6Eg8tBJ5vT"}]}]}} {"custom_id": "Jhenaidah_23598_6156_24111_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3pQUKSUpZgQ91RDJAFc9nf"}]}]}} {"custom_id": "Sunamganj_13338_19494_13851_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XmbYPiN6fRWHCFXNceywLS"}]}]}} {"custom_id": "Faridpur_21546_13338_22059_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ax7y6naT1Gqy1AjYX9ZidC"}]}]}} {"custom_id": "Pabna_23598_22059_24111_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HSFscjn7a9umMZfiiag8yL"}]}]}} {"custom_id": "Khagrachhari_17955_37962_18468_38475.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VrJdgbb966aAFJPUczgfPn"}]}]}} {"custom_id": "Kurigram_3078_12825_3591_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XEooWMjou7zt3RKa4vvqTK"}]}]}} {"custom_id": "Pirojpur_7695_20007_8208_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1zQgciPPDWrWFRgmHb9Jb9"}]}]}} {"custom_id": "Jhalokati_2052_22059_2565_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7brvjq8xMobSpS8zmsMnbv"}]}]}} {"custom_id": "Barguna_7182_13338_7695_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EXNMdyoy9vCHwa36yBejTj"}]}]}} {"custom_id": "Barisal_15903_4104_16416_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Y3d5t1DMLbWs47hpj5qoDR"}]}]}} {"custom_id": "Rangamati_22059_46683_22572_47196.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1fKnsZoBrqNgrh58WuGYgD"}]}]}} {"custom_id": "Sylhet_6156_2565_6669_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YYBUbr5xyE4V1SuL2PyX2v"}]}]}} {"custom_id": "Panchagarh_13851_26163_14364_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9qkrvqdAekdA8JuQcGA9C9"}]}]}} {"custom_id": "Kishoreganj_12825_8208_13338_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SFbsD3AJcT8XJgwzbvYWSv"}]}]}} {"custom_id": "Bandarban_4104_7182_4617_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SAAnetVk6UCnwhm9g8wVDc"}]}]}} {"custom_id": "Sylhet_15903_26163_16416_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1X1brJ1Wbv3qJqkevHzY6N"}]}]}} {"custom_id": "Sylhet_8208_17442_8721_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8KCrxwi6X1fSkv7VmNytzD"}]}]}} {"custom_id": "Rangamati_30780_74385_31293_74898.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UbMvvFn8DtR567yjM2Uehp"}]}]}} {"custom_id": "Pirojpur_7182_22572_7695_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Dv6F7GmoHv9vSQUjDEqsiz"}]}]}} {"custom_id": "Sirajganj_13338_16416_13851_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TjHWKE76Uf9BDRX9xEePAJ"}]}]}} {"custom_id": "Pirojpur_12312_18981_12825_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SFJpqfWPnkbh2EzJ92MPyC"}]}]}} {"custom_id": "Maulvibazar_16416_11799_16929_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TeX7df2WsFm87X5NVT5ay5"}]}]}} {"custom_id": "Habiganj_20520_12312_21033_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5Nuf4uUVai412vVvnh8q2H"}]}]}} {"custom_id": "Shariatpur_15903_6156_16416_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SXwVuVBuGNWGC5oz4wGuoJ"}]}]}} {"custom_id": "Lalmonirhat_19494_27189_20007_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9Afz4xn6xaMYvhtNUQa8HY"}]}]}} {"custom_id": "Jessore_7695_12312_8208_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K4jCPtcGUPeqCcfcoocPKV"}]}]}} {"custom_id": "Lakshmipur_10260_21033_10773_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VtNeuLgxb65Y7imkXWuwwE"}]}]}} {"custom_id": "Lalmonirhat_1539_513_2052_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NgQbSPDRkUmhoFURhCgGeD"}]}]}} {"custom_id": "Jamalpur_6669_10260_7182_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PEyWE4KmjR5dejNvJ8r9Vy"}]}]}} {"custom_id": "Satkhira_8721_23598_9234_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5JP41z7ZFZGq3oBQpnKjJ4"}]}]}} {"custom_id": "Joypurhat_1539_2565_2052_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7vD8odHQfpMBXePXch6fdP"}]}]}} {"custom_id": "Jessore_6156_17955_6669_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8q63gfzNeMkHpPDBkusc2C"}]}]}} {"custom_id": "Lalmonirhat_17442_27189_17955_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LprVvHsiVAQcopmbx3BrSH"}]}]}} {"custom_id": "Sherpur_9747_15903_10260_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TE1tQBhHiDUUaBTrwHHWJk"}]}]}} {"custom_id": "Jamalpur_10260_27702_10773_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N3ucvmVRSsFoBqf3jTAYvt"}]}]}} {"custom_id": "Bhola_22572_32832_23085_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LHJq9z5Co8eYCRyEHKGkK9"}]}]}} {"custom_id": "Manikganj_15903_11286_16416_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FMtauighaHHnqp1ubTbmj8"}]}]}} {"custom_id": "Khulna_1026_9747_1539_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UDgpmQ3dvHZWg3n4JH55DX"}]}]}} {"custom_id": "Thakurgaon_14877_4104_15390_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5zgus7SE1iVwjKP19ybqUT"}]}]}} {"custom_id": "CoxsBazar_11286_21033_11799_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F5nWAYWZ3oQeopgxtp8PJn"}]}]}} {"custom_id": "Kushtia_5643_10260_6156_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CN33m2qPjdBsWbKR2F7gGB"}]}]}} {"custom_id": "CoxsBazar_10773_29241_11286_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GyzUTdsgEzmmCdZP2CQqT7"}]}]}} {"custom_id": "Sylhet_10773_15390_11286_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-42AV4bQL2WrQDs6g2nPCKK"}]}]}} {"custom_id": "Gaibandha_1026_23085_1539_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bwftgaji1wahcvL6X44Cqp"}]}]}} {"custom_id": "Joypurhat_8208_19494_8721_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5Gf5yojGoC6vXfAa33reKF"}]}]}} {"custom_id": "Mymensingh_25650_35397_26163_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YbvkJUHQmZRCgBN6PVyZC8"}]}]}} {"custom_id": "Joypurhat_2565_3078_3078_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-USpLsfi5uAncSefAUvoqnA"}]}]}} {"custom_id": "Satkhira_6669_30267_7182_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QEaHVCRMTeF71YmPitZmAH"}]}]}} {"custom_id": "Bandarban_23085_47196_23598_47709.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LyDc4J6CFpkbLe4hZ6k8uy"}]}]}} {"custom_id": "Barguna_16416_17442_16929_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7NkrzmfEtmtwuainbH5Y6G"}]}]}} {"custom_id": "Jhenaidah_13851_9234_14364_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8UyszjsvnSQQB7us2Nj5Dw"}]}]}} {"custom_id": "Lakshmipur_9747_7695_10260_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D38bfyw1tdDF96j4kUwR8n"}]}]}} {"custom_id": "Rajshahi_21033_13851_21546_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M5d8zdt2cpTqhkQhMdhdDJ"}]}]}} {"custom_id": "Narail_8721_7182_9234_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SEgSFtErYhdCNjMouHHRRS"}]}]}} {"custom_id": "Maulvibazar_29241_1539_29754_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HrvQppAVXWNwf6MtrEVxdi"}]}]}} {"custom_id": "Jessore_13851_21033_14364_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ku4iAvksNEVUwKrHdX82aJ"}]}]}} {"custom_id": "Chittagong_36423_28728_36936_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LnVdMBm6VgN4aS7FEAqwnQ"}]}]}} {"custom_id": "Jessore_12312_15390_12825_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XLinpJwb4JFDMhSBQVNFuo"}]}]}} {"custom_id": "Maulvibazar_7695_23598_8208_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3mWzhYQfuM4UtG4ah4Y1aT"}]}]}} {"custom_id": "Noakhali_2565_14364_3078_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5sGbKTCEHHtJharx9Y76cf"}]}]}} {"custom_id": "Jhenaidah_25650_6669_26163_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U3DsF46ibe1VysyxoKt4gU"}]}]}} {"custom_id": "Kishoreganj_25137_11799_25650_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9gcDEunVAw6hwpsnnseqH9"}]}]}} {"custom_id": "Habiganj_16416_17442_16929_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fkgjg8dPsh2GZ5VRkgkGm1"}]}]}} {"custom_id": "Rangpur_17955_16929_18468_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MwQQjCefN3UB4S5pcXKnk4"}]}]}} {"custom_id": "Gaibandha_10773_11799_11286_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VwGdTW4DTa28iNYVdCssop"}]}]}} {"custom_id": "Mymensingh_28728_29241_29241_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SFFGt5mJMDd9Ge6c5pCfjW"}]}]}} {"custom_id": "Khulna_10773_18468_11286_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FDkuPfQuSEJKiWhTJPskEt"}]}]}} {"custom_id": "Manikganj_15903_15903_16416_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QxHtLeH5kNRkWab4WLDvHi"}]}]}} {"custom_id": "Barisal_5643_4104_6156_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-McSFDYKUfr2hkQDdrZA3tg"}]}]}} {"custom_id": "Maulvibazar_8208_18468_8721_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2AyTgmrjuHRUM92d5HwdFt"}]}]}} {"custom_id": "Sylhet_27702_13338_28215_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-E4p9Cr3TCpo7SXFAJBh2GT"}]}]}} {"custom_id": "Mymensingh_13338_33858_13851_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5wCdFCnAWCnXz6MT6JuiDR"}]}]}} {"custom_id": "Naogaon_7182_1539_7695_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M2HeZHoeeVMosza9496R5m"}]}]}} {"custom_id": "Jamalpur_7695_513_8208_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5rVNQhL9oLMg99ceyoWm2Y"}]}]}} {"custom_id": "Chittagong_12825_14364_13338_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-16QE4zggTgLZSBCgmrHq3e"}]}]}} {"custom_id": "Habiganj_18468_3591_18981_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CADapZG49empUfYaWaknS8"}]}]}} {"custom_id": "Manikganj_1026_8721_1539_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MDED8gee2xpESMdxmdhtuE"}]}]}} {"custom_id": "Sherpur_17955_10773_18468_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-69CJpRACaaatJ8Wa9SQ1dK"}]}]}} {"custom_id": "Noakhali_4104_14364_4617_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PGryhryhXoBKHun4z1vRxN"}]}]}} {"custom_id": "Sirajganj_16416_13338_16929_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VUgEiiSUqGN8SMcAGoGR6U"}]}]}} {"custom_id": "Comilla_17955_8721_18468_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4YMrqzC2P34gjNssXN9JqN"}]}]}} {"custom_id": "Nawabganj_5643_17442_6156_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LbY2ALBrbJH7uu6h4j4Rna"}]}]}} {"custom_id": "Rangamati_21033_36936_21546_37449.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-93pUWRArqhnxV4EQ2kViJa"}]}]}} {"custom_id": "Brahamanbaria_24111_3591_24624_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A49xaCPvJpnRaRb3kNqHnk"}]}]}} {"custom_id": "Naogaon_22059_22059_22572_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A7gLqrBE6nt6g4iZwp34Kx"}]}]}} {"custom_id": "Barisal_22059_2052_22572_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TSVMdZ22tDHyTtptmWeJrp"}]}]}} {"custom_id": "Rajshahi_23085_3591_23598_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2Dxp2EzxcspPXKSUBcgHpG"}]}]}} {"custom_id": "Chittagong_25650_19494_26163_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ALDhjAUZPP6vKCr6S13kMN"}]}]}} {"custom_id": "Chittagong_31806_47196_32319_47709.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KXx3wvWd7LZnbYTjf2ig8G"}]}]}} {"custom_id": "Barisal_10773_29754_11286_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MXGMU3zVdnMFHehMLDzt4P"}]}]}} {"custom_id": "Kurigram_14877_31806_15390_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EtFTpFkcrqpdQaFq9QGroZ"}]}]}} {"custom_id": "Manikganj_16929_15903_17442_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M4iEUyhZ1D3DnkoFr4FMkj"}]}]}} {"custom_id": "Bandarban_6669_48735_7182_49248.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-75UH88H8eBpXLbh8fNHyHh"}]}]}} {"custom_id": "Rangamati_13338_38475_13851_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-54XnN3raXFLp2Req8UHEaq"}]}]}} {"custom_id": "Dinajpur_10260_23085_10773_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QT5eHh2P7xAozdCpz6TZdZ"}]}]}} {"custom_id": "Joypurhat_4104_5643_4617_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VcTXao2iMfWrBQeR4H7iPN"}]}]}} {"custom_id": "Mymensingh_17442_26676_17955_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8tMC1iSPDN6Hc3zgK1Ki7a"}]}]}} {"custom_id": "Habiganj_12825_7182_13338_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JfUu598iHLu2TUC52XsaPe"}]}]}} {"custom_id": "Brahamanbaria_2052_27702_2565_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TPd9Zs48MSBiGFxLojhmjN"}]}]}} {"custom_id": "Nilphamari_1539_16416_2052_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-62B8RAWTHnMQDWP4CtzxcX"}]}]}} {"custom_id": "Bandarban_8721_26676_9234_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Qb8qVMC4vsjzjM79DX6Nt1"}]}]}} {"custom_id": "Sunamganj_31293_14364_31806_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U4w1wuziofJUMYM7ALExE4"}]}]}} {"custom_id": "Gaibandha_12312_3078_12825_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WicBnCkHqCqLSqvRkPhBmn"}]}]}} {"custom_id": "Kurigram_8721_9747_9234_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PCxykyQQ7jHsL8fHWa29CA"}]}]}} {"custom_id": "Bhola_8208_36423_8721_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UvX4nM5xVfnAoM8fc2QG7q"}]}]}} {"custom_id": "Maulvibazar_20520_15390_21033_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MniufpXh9vuqMNKdQFa2Sw"}]}]}} {"custom_id": "Sylhet_8208_27702_8721_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D7Pc9xS3CgWoqw4Gaawwz6"}]}]}} {"custom_id": "Barguna_5130_17955_5643_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1hm3ahZhpmmKvRFuQgDTfi"}]}]}} {"custom_id": "Sherpur_5130_17955_5643_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4n9daVZdqJjZD44SYbL7UY"}]}]}} {"custom_id": "Faridpur_8208_14364_8721_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AJRpeT39hnXyvz312keF4u"}]}]}} {"custom_id": "Dinajpur_16416_20007_16929_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TAFLKf9DWPxuxCUm5ZVGUt"}]}]}} {"custom_id": "Bagerhat_20007_24111_20447_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Cm4pQCNP4EivT6sHU6cVYb"}]}]}} {"custom_id": "Chittagong_25650_32319_26163_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TErMwni2ro2o2uQgVJNykw"}]}]}} {"custom_id": "Chittagong_30267_27702_30780_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GjvFk5V6AXBRfBCsVfpRPb"}]}]}} {"custom_id": "Comilla_28728_24624_29241_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7dTLsGRgHZzgFsVjNoEaur"}]}]}} {"custom_id": "Chittagong_22059_10260_22572_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JqY3FDkqGiarsoY7WgPXy7"}]}]}} {"custom_id": "Chittagong_32832_49248_33345_49761.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TZErHZKrWphySHsGXH2yXN"}]}]}} {"custom_id": "Maulvibazar_1026_31293_1539_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-THgcCfS2dgg8cmymG9iezf"}]}]}} {"custom_id": "Jhenaidah_15903_6669_16416_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TSYUQVYZjW2mcoiutBza8H"}]}]}} {"custom_id": "Khagrachhari_10260_48222_10773_48735.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YApWzdJHorg3AyCHyaQqHM"}]}]}} {"custom_id": "Sirajganj_12312_18468_12825_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Eg5oxVApuU7UGFATLNM75j"}]}]}} {"custom_id": "Joypurhat_11799_12825_12312_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6BQyYon6tYZDKNGH5d2oet"}]}]}} {"custom_id": "Feni_8208_8721_8721_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7T1QUFzKnAK3Vetd1CW9fw"}]}]}} {"custom_id": "Gopalganj_20007_12825_20520_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BtkYPAEJ31XkBnUPtzL845"}]}]}} {"custom_id": "Natore_4104_14877_4617_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7oN1TbW97j1Gx4EiNqfa4q"}]}]}} {"custom_id": "Bagerhat_11286_2052_11799_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JT6Jgvqdr4ihCUnzg2ws2J"}]}]}} {"custom_id": "Barguna_15903_1539_16416_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GBr2xrs4ojzUg1RyVjeh6V"}]}]}} {"custom_id": "Satkhira_18468_43605_18981_44118.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SJUyGbdAQM1j8hfDnaxtGE"}]}]}} {"custom_id": "Rangpur_9747_16929_10260_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VY2WTDfHxfWruCEEax8qg3"}]}]}} {"custom_id": "Narail_9747_5130_10260_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9AwRTvbZieVnN32zW2MUsz"}]}]}} {"custom_id": "Sylhet_5643_15390_6156_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EBoWGxHhYLKyGmL52vzwmk"}]}]}} {"custom_id": "Jhenaidah_19494_24624_20007_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7t7jsCbeqJx8rADwZMyXN7"}]}]}} {"custom_id": "Rajshahi_22059_5643_22572_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5y4QEbVbqiQqmrG2os4LFg"}]}]}} {"custom_id": "Narsingdi_5643_10260_6156_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NwNHFjCNTzrsWGUSznK1pK"}]}]}} {"custom_id": "Netrakona_18468_7695_18981_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PMXMa4XFTgPiFu3PPRpzXn"}]}]}} {"custom_id": "Chuadanga_8721_20007_9234_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M89SKFdsNERUmroFtu9LPq"}]}]}} {"custom_id": "Noakhali_3078_5130_3591_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NSKUHXBPpwoxkZGP5V5bFv"}]}]}} {"custom_id": "Rangpur_7695_1539_8208_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S2B3YcxxGGcay52mYdrmPS"}]}]}} {"custom_id": "Gaibandha_13338_4617_13851_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9ujFTUT8dDqmqnjfnXj8s4"}]}]}} {"custom_id": "Habiganj_11799_22572_12312_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vf3G7qwCcHNy9mNhGmioPb"}]}]}} {"custom_id": "Chandpur_7695_7695_8208_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XnLw8Hd7QeUyqQsSuMVmsj"}]}]}} {"custom_id": "Kurigram_12312_34371_12825_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-93kPB7chjfje82Hy9h5uvT"}]}]}} {"custom_id": "Habiganj_20007_21033_20520_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Au1S2Ze7xgoE4DLY42RXd1"}]}]}} {"custom_id": "Tangail_21033_20007_21546_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WQYJkJrSWi7KHUGRn5grPp"}]}]}} {"custom_id": "Khagrachhari_9747_34371_10260_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NWrKzRcKiXQnXgyJ2K8u8Z"}]}]}} {"custom_id": "Noakhali_3591_31806_4104_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D7vgQTCpjh7TX7LQJyfHfC"}]}]}} {"custom_id": "Gaibandha_10260_3078_10773_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RKYSeeRP9AcknBmj1jUAk1"}]}]}} {"custom_id": "Kurigram_12825_9747_13338_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7HGs1bMfaVJRjUo1LwwSCU"}]}]}} {"custom_id": "Rangamati_16416_25137_16929_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fd58EwZ9jtudd25ZxHjWe8"}]}]}} {"custom_id": "Nawabganj_16416_14364_16929_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9JYdjK7ZxRspdQkeEurjB6"}]}]}} {"custom_id": "Bogra_30267_5643_30780_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FBxNhBRANmegwBUiNQhNgb"}]}]}} {"custom_id": "Jhalokati_7182_10773_7695_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Spg7bjduq4ttxuw2nTtpA3"}]}]}} {"custom_id": "Bandarban_4617_38475_5130_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Nv1DusAR95fZdBFZm4Q3MG"}]}]}} {"custom_id": "Bandarban_6156_54378_6669_54891.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2LRSHEvTSEdFrGYjAchLkh"}]}]}} {"custom_id": "Kushtia_25650_22059_26163_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VN9bq2cJK2uXkbqQ4tXvMZ"}]}]}} {"custom_id": "Rangpur_27189_16929_27702_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U6fZxjnS6BynNU3tjLegRZ"}]}]}} {"custom_id": "Barisal_10773_9234_11286_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BYdMyxq7C67Psm2aAXE7bB"}]}]}} {"custom_id": "Khulna_13338_24624_13851_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8tr5n6NXPCjcDgjUyV1dfm"}]}]}} {"custom_id": "Noakhali_3591_7695_4104_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5wMQ1nbfUb54pmkPC8DrSp"}]}]}} {"custom_id": "Lakshmipur_6156_16929_6669_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FD5bKLSQaW7qUnErj6zFdU"}]}]}} {"custom_id": "Dinajpur_30267_31806_30780_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VnzqsGLCiADLjG4DMK8Cu9"}]}]}} {"custom_id": "Natore_10260_12825_10773_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LxgkhuQHgKsEVkRBMVb7bz"}]}]}} {"custom_id": "Narsingdi_5643_6669_6156_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fqy1RDCXSjEekv4iAWR2Cu"}]}]}} {"custom_id": "Rajshahi_15390_8208_15903_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D4KJ28F5ztC52Cabax38vy"}]}]}} {"custom_id": "Bagerhat_13338_19494_13851_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7fW6Jf8PQdNJFLZLpMNEiz"}]}]}} {"custom_id": "Bandarban_4617_2052_5130_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Prq87uTqbFWu45qvSZgy5B"}]}]}} {"custom_id": "Sylhet_5643_29241_6156_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XwnGE1cd5r1Yowizmn9eBj"}]}]}} {"custom_id": "Rangamati_1539_55917_2052_56430.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Lk8g7u2jiJCutwxyE7uLQ5"}]}]}} {"custom_id": "Thakurgaon_5130_13338_5643_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8adYPkgBAAQzd39MQbCaDa"}]}]}} {"custom_id": "Kushtia_18468_21033_18981_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RtJ31YQv9nVgWjokszEYVp"}]}]}} {"custom_id": "Kurigram_13851_16929_14364_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XACb7B2gd2FDWNGcnT5dwt"}]}]}} {"custom_id": "CoxsBazar_4104_20520_4617_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-29RFAgFqTAHm9b97QDeTmr"}]}]}} {"custom_id": "Mymensingh_5130_21546_5643_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bc3qRrNrQp54QJ4WEdfF4Q"}]}]}} {"custom_id": "Gaibandha_13851_4617_14364_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WVquBM9sQjf8TeZ3t1QXsV"}]}]}} {"custom_id": "Pirojpur_5643_4104_6156_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FDD3qPUJQ1czhno8Luvgvh"}]}]}} {"custom_id": "Kishoreganj_20007_15903_20520_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MKTHC8ZU1jcxtmNqqP5R7W"}]}]}} {"custom_id": "Rangamati_27189_81567_27702_82080.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GYR8maURrEh5BtRKSaRRkJ"}]}]}} {"custom_id": "Netrakona_26163_17955_26676_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CNPs3ej7QYrn8Eppu81EVu"}]}]}} {"custom_id": "Faridpur_8208_6669_8721_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LXQKVVp1vqMDnUKYx9koTE"}]}]}} {"custom_id": "CoxsBazar_12312_36423_12825_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-33ecoqumv3BSNCVmhpkQHT"}]}]}} {"custom_id": "Noakhali_8721_42579_9234_43092.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SbZ6Mmw2S7jTrc5yrRjGNe"}]}]}} {"custom_id": "Jessore_13338_26163_13851_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-93Q5cM4FwFN6ewt5K8R4Vp"}]}]}} {"custom_id": "Maulvibazar_13338_21546_13851_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PGG9vvShg81jmfn6cN6aSk"}]}]}} {"custom_id": "Noakhali_19494_26676_20007_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1N3449qetwc54K1i6TCSzq"}]}]}} {"custom_id": "Madaripur_12825_10260_13338_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W6n1vzspfercjgq9dZJX4G"}]}]}} {"custom_id": "Natore_13338_2052_13851_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-STPK21GDNXepbVQH9Tqozb"}]}]}} {"custom_id": "Patuakhali_14877_13851_15390_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ERXecEoneAmxRcmDFYA45h"}]}]}} {"custom_id": "Brahamanbaria_17955_15390_18468_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QuqLuzeQBX52EZpGvLrvGK"}]}]}} {"custom_id": "Manikganj_24111_14877_24624_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ShpvHY26jf729iZDum3ZqX"}]}]}} {"custom_id": "Bhola_20520_30267_21033_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MZm3JbmDVT7o16cfQqWjvo"}]}]}} {"custom_id": "Sylhet_14877_16929_15390_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-57hfbMjZaH5Ddz2kzjrAsz"}]}]}} {"custom_id": "Rangamati_10260_64125_10773_64638.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LEKGoujwnbyRAXmjj7txhM"}]}]}} {"custom_id": "Mymensingh_5643_33345_6156_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-C44mZTWd4vQnoFW8eK3S3V"}]}]}} {"custom_id": "Bhola_10773_28728_11286_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-De3TkY5DSjcrFaZm61ASaM"}]}]}} {"custom_id": "Dinajpur_24624_34884_25137_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VCuhQMc8anEjTUkZQCW971"}]}]}} {"custom_id": "Mymensingh_7695_17955_8208_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-87tyYf13eC9V6n9azBpKzf"}]}]}} {"custom_id": "Maulvibazar_4617_28215_5130_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PkqCN3jHZf4P5LRVdygqXE"}]}]}} {"custom_id": "Rangpur_17955_6156_18468_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1bKaSJ5NExnFzmXtFLFw7E"}]}]}} {"custom_id": "Khulna_6669_42066_7182_42579.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-71Jh5VWeKH2gEEdQRFP7Bp"}]}]}} {"custom_id": "Tangail_17955_35397_18468_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6ui1sMJLpuXusHYweGYQYD"}]}]}} {"custom_id": "Gaibandha_12825_27702_13338_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SL48B6Q1YxaLga6Tc45qnX"}]}]}} {"custom_id": "Kurigram_17442_26163_17955_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FxFed6Xh4N4VNBz6UvWDjt"}]}]}} {"custom_id": "Noakhali_16416_25137_16929_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EncmqFSKFqwsm7wKv5rHd2"}]}]}} {"custom_id": "Feni_10260_9747_10773_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8sVDZCS9xq6HpHRkrFCnXX"}]}]}} {"custom_id": "Jamalpur_2052_23598_2565_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WpBadANDbe1rykPaonoX32"}]}]}} {"custom_id": "Rangpur_24111_18468_24624_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Cahj7yvXamAiRtZQYFYUob"}]}]}} {"custom_id": "Habiganj_14877_26163_15390_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L3esGEBY7L4Uzci1Emkgkc"}]}]}} {"custom_id": "Netrakona_21033_27702_21546_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vq15cUQemdAxrN6Ek35qad"}]}]}} {"custom_id": "Narsingdi_8721_21033_9234_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7dCSW6aeY2bwFHvPHQs1ZA"}]}]}} {"custom_id": "Khulna_5643_23085_6156_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HRxtTa5cy13zMXAbxxdZDU"}]}]}} {"custom_id": "CoxsBazar_14364_7182_14877_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bt5AwQ2vsNnb2DRgcRQiLw"}]}]}} {"custom_id": "Sirajganj_20007_27189_20520_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3mKUJ3BqxWYbvWn3BNkS88"}]}]}} {"custom_id": "Jessore_14877_8208_15390_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Uk5m73uYxiQm4eNbhCt6rv"}]}]}} {"custom_id": "Kushtia_5130_6669_5643_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SSyYJ85oP354jymLbPEZam"}]}]}} {"custom_id": "Brahamanbaria_18468_25137_18981_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8UvYNAry9Lw6u9C3ZS83hA"}]}]}} {"custom_id": "Habiganj_18981_2052_19494_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9PKGqVD4xHsGGFvQUFMjL1"}]}]}} {"custom_id": "Noakhali_7695_48735_8208_49248.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D786dBiRXsQnCZWHZiUocC"}]}]}} {"custom_id": "Barguna_17442_20007_17955_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QNJkATKwAmWAsAGYW55jkx"}]}]}} {"custom_id": "Rangamati_9747_69768_10260_70281.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Qh6fjLtKZqtyuhUhuiaMNz"}]}]}} {"custom_id": "Bagerhat_2052_53865_2565_54378.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CiPJJ74jrPciEjqTA6z4fW"}]}]}} {"custom_id": "Bogra_17955_8208_18468_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CCiQtcHaMCSfCoXtp6q8zE"}]}]}} {"custom_id": "Chuadanga_5130_14364_5643_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6xrZdt2cBgAzrahxj18pS8"}]}]}} {"custom_id": "Narsingdi_7182_3078_7695_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K3Bz5fz63oMi1AT229F1dP"}]}]}} {"custom_id": "Patuakhali_20007_34371_20520_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Di1kDyPx5rJzEd6tovdxpL"}]}]}} {"custom_id": "Panchagarh_18981_19494_19494_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MyMvdwM2hFxcaCzkpNUvru"}]}]}} {"custom_id": "Feni_8721_8208_9234_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6bnr1yywLFpLvEbhB6gg45"}]}]}} {"custom_id": "Chittagong_25137_40527_25650_41040.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ds8N1AV6mDzgx8r7uPpbSZ"}]}]}} {"custom_id": "Sunamganj_14364_2565_14877_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A3hTUn9YnLvqThiWRQoiCj"}]}]}} {"custom_id": "Joypurhat_11286_2565_11799_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6kGzHvDgsKhJ5anHrKwgNv"}]}]}} {"custom_id": "Jamalpur_7182_34884_7695_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YQaJJ7NYrYsuwPcqcq6zU2"}]}]}} {"custom_id": "Rajshahi_10773_8208_11286_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9tpt4kAzGTKpo6vQvKYAPn"}]}]}} {"custom_id": "Dinajpur_12825_30267_13338_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7eqmZZ32LtmFftKe7macDy"}]}]}} {"custom_id": "Comilla_17955_21546_18468_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DXMfZm2wGbufZE7UkJoxXG"}]}]}} {"custom_id": "Comilla_6669_3078_7182_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-96TVBtrXWmwpDGE3tDVTTg"}]}]}} {"custom_id": "Brahamanbaria_25137_5130_25650_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6qajHYwnnP4m4YpHYUWbAw"}]}]}} {"custom_id": "Magura_4617_12825_5130_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LNZSrpB5Ho7ZMJwVLLXgWo"}]}]}} {"custom_id": "Lakshmipur_10260_15390_10773_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RHncWKxJdHX5srwa3e5ppp"}]}]}} {"custom_id": "Rajshahi_22059_5130_22572_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PmaEckNuxR5Af6gJLR7WEr"}]}]}} {"custom_id": "Narayanganj_6156_3078_6669_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PXz9fnJKrz7hbXsxuXK7GR"}]}]}} {"custom_id": "Comilla_26163_22059_26676_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CtV3xvcyTriPhX3LHgehZF"}]}]}} {"custom_id": "Bhola_6669_16929_7182_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3h4VYRubKwy93WKbfxppZo"}]}]}} {"custom_id": "Rajbari_20520_10260_21033_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TzPWYsa2396kHGjhWHDWNt"}]}]}} {"custom_id": "Gazipur_10773_9747_11286_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ck4Bp3pgz7EKNGNSx5ywom"}]}]}} {"custom_id": "Narayanganj_3591_5643_4104_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NJ3u7hFWDMEYNEjHC3rAsQ"}]}]}} {"custom_id": "Khagrachhari_11799_35397_12312_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3SyyaUQLpY3Td7HothtThF"}]}]}} {"custom_id": "Kushtia_12312_7695_12825_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VraWsCMsG1wKujkTNgJqHT"}]}]}} {"custom_id": "Bogra_1539_15390_2052_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9Nb8F5mQjaTcsrM34V9myo"}]}]}} {"custom_id": "Jessore_17955_16416_18468_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2mKuTzx89X3c1DRPCE6vJ4"}]}]}} {"custom_id": "Thakurgaon_10773_17955_11286_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WEM2aFV6GHV9NPDv87R37i"}]}]}} {"custom_id": "Khulna_6156_62073_6669_62586.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MWigcKz4VqpvxjtwaHfxQ5"}]}]}} {"custom_id": "Chittagong_37449_49761_37962_50274.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SPUH3oaYdM4XuJURMUUudr"}]}]}} {"custom_id": "Jhenaidah_26676_8721_27189_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R4NGKdwHYwQQrDpBsJZKVx"}]}]}} {"custom_id": "Sylhet_23598_6669_24111_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SF6Dk9Q7pXY7roAXbMCkq4"}]}]}} {"custom_id": "Nawabganj_8208_25650_8721_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5gLpL9cRDxUuMa6Z3Pm1nm"}]}]}} {"custom_id": "Noakhali_9234_16416_9747_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-33ZcoDPpVy7hmXFcJLids4"}]}]}} {"custom_id": "Dinajpur_20007_17955_20520_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FGct4kKrXgzJki3GfwPnNz"}]}]}} {"custom_id": "Jamalpur_14364_29241_14877_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F1ZKgE43tWpJ9oytLo3bj3"}]}]}} {"custom_id": "Gazipur_13851_16416_14364_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XJ89Ybu3UWDFBBwYUXbiVL"}]}]}} {"custom_id": "Sherpur_7695_14364_8208_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JB5vgEYFB8JXE3hS48Rfv7"}]}]}} {"custom_id": "CoxsBazar_13338_27189_13851_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QVEDvW2aCDAYK7vgAroid2"}]}]}} {"custom_id": "Shariatpur_0_2052_513_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VDK6tFKKY8ySNtpqoGHtRy"}]}]}} {"custom_id": "Gazipur_18981_22572_19494_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Big6Q1vgfDJgFX9dPWQRzX"}]}]}} {"custom_id": "Bagerhat_8208_43092_8721_43605.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8n8eEiwW6SDtmMLbHBnPx4"}]}]}} {"custom_id": "Comilla_24624_15903_25137_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Xdx7fPzrG3iasHq9MYHYKN"}]}]}} {"custom_id": "Gopalganj_12825_15390_13338_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-E3ZXG4VHB6BYKgp2Nf4Cs5"}]}]}} {"custom_id": "Rajbari_2565_7695_3078_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EqKJ9AYYhnJFv4hZpDVNk3"}]}]}} {"custom_id": "Sirajganj_15390_15903_15903_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JJrAop7XipzAjbujrrYc2J"}]}]}} {"custom_id": "Nilphamari_8208_17442_8721_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RGQHJWKfTtNypL7yQQamjP"}]}]}} {"custom_id": "Kushtia_25650_13851_26163_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3e46PgnpHEMnE2ZKUVdRK8"}]}]}} {"custom_id": "Madaripur_6669_12825_7182_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FmuEMkyJzjMboSavKimfyS"}]}]}} {"custom_id": "Tangail_18468_17442_18981_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V7KNyHjxVFdGZhRkTDvx48"}]}]}} {"custom_id": "Jamalpur_10773_10773_11286_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LNM4D1vcphLRzfb81q86SX"}]}]}} {"custom_id": "Faridpur_25650_11799_26163_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LtCnL3dWNYVLEaQuYPwLJA"}]}]}} {"custom_id": "Rajbari_1026_5643_1539_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WQicCKwcdeGEnTJ2rBUD3s"}]}]}} {"custom_id": "Barisal_23598_4104_24111_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HmJmykEGGYpzzmaaqGgGGa"}]}]}} {"custom_id": "Satkhira_10260_9234_10773_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YLQWQSfiFGVJ8vPJQE2BNz"}]}]}} {"custom_id": "Rangpur_11286_10260_11799_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XYj13tV5mzemiLhwjeiXeJ"}]}]}} {"custom_id": "Mymensingh_10773_24624_11286_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CdxvWiwHs2ziAxpJ6U8VBZ"}]}]}} {"custom_id": "Jhenaidah_30780_6156_31293_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6RezD75mgdPy8sXZtkuUaN"}]}]}} {"custom_id": "Bandarban_21546_44118_22059_44631.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XSVgbBVzAqz2JCeYEeJz5e"}]}]}} {"custom_id": "Faridpur_12825_18468_13338_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GeQDLkAXpe82bL9MebQ6qw"}]}]}} {"custom_id": "Kishoreganj_28215_13851_28728_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D7Y48Abh6CqmhgYoj3MNa5"}]}]}} {"custom_id": "Faridpur_23085_9747_23598_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XdxrWdLNjqEwMb3K7gQDuP"}]}]}} {"custom_id": "Bandarban_13338_10773_13851_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-79TSoKcNVbvfML4DqKicJd"}]}]}} {"custom_id": "Bandarban_8721_22059_9234_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PAYezJCVfksC7FBff7aNw6"}]}]}} {"custom_id": "Rangamati_17955_31806_18468_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UMp9cv9MxqzLTYpNfqam4o"}]}]}} {"custom_id": "Bagerhat_14877_51813_15390_52326.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N6kEEGQyowyo8LzAK79KbQ"}]}]}} {"custom_id": "Sylhet_34371_9234_34884_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WGZJQBdDNLsNyjUVTEPcHR"}]}]}} {"custom_id": "Bogra_25137_5130_25650_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6B4z6ELPyFBskNpWhC1Amk"}]}]}} {"custom_id": "Pirojpur_8208_24624_8721_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AYeUfEUiKUMeofgdeinL8a"}]}]}} {"custom_id": "Khagrachhari_5130_44631_5643_45144.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LskuGFm6Co3fqhwYWgSMWR"}]}]}} {"custom_id": "Nilphamari_17442_17442_17955_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UJsWC4dFGWPhCssRwuChiT"}]}]}} {"custom_id": "Lakshmipur_9747_3078_10260_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K5r3NzSgamEHT7BdH4rRhV"}]}]}} {"custom_id": "Sherpur_13851_13338_14364_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-92eMpN2quhvvkVL9E8W14L"}]}]}} {"custom_id": "Comilla_2565_15903_3078_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Wje4ydiEi5aZSmXp3N5XMY"}]}]}} {"custom_id": "Dinajpur_22572_24111_23085_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6Sj6BZKBz8tceJTFWriFHp"}]}]}} {"custom_id": "Tangail_17955_13338_18468_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ssn93KsLCHJ52zLyu3MCwb"}]}]}} {"custom_id": "Chandpur_17955_7182_18468_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RvFRBC5D3WnjyS9A64v1Ln"}]}]}} {"custom_id": "Netrakona_16416_29241_16929_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GsMu7xXQAsnpgr4VVGR3f7"}]}]}} {"custom_id": "Bogra_15903_17442_16416_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D41KV39mKM8SJ4pBaaYvx2"}]}]}} {"custom_id": "Jhenaidah_7695_21033_8208_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CYKGheHgFHddtBRRTvtsSK"}]}]}} {"custom_id": "CoxsBazar_8208_11799_8721_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BCpEgeH8jewd2k2iixw6DW"}]}]}} {"custom_id": "Barisal_16929_12312_17442_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BnnypjNZeyU3DMA74W3pMW"}]}]}} {"custom_id": "Sylhet_14877_17955_15390_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NZvDroAr7TSsN9YCRkg8tG"}]}]}} {"custom_id": "Panchagarh_4617_20520_5130_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WtUNqzmxLZ3qEp78Nckyy1"}]}]}} {"custom_id": "Patuakhali_2565_32319_3078_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QmHWt67JEqcVH496mPtGFe"}]}]}} {"custom_id": "Bhola_6669_26163_7182_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V8omssZK3Lapw51NVBMEgH"}]}]}} {"custom_id": "Faridpur_20007_20007_20520_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5FNb6CY9m4ceant9DNMKqu"}]}]}} {"custom_id": "Bagerhat_4617_25137_5130_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Y53tkci8ir2S3cmmw7Gud9"}]}]}} {"custom_id": "Feni_7695_2565_8208_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2aDfusLsuXLbUdWm8UtE5j"}]}]}} {"custom_id": "Rangamati_9747_47196_10260_47709.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Dnu7bUegxjNkhQ9XfKhyuj"}]}]}} {"custom_id": "Thakurgaon_10260_15903_10773_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XGhnTNnUmMA1e6aGMTqMjr"}]}]}} {"custom_id": "Netrakona_19494_7182_20007_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8KewTfW7MzsavcPnYbSezg"}]}]}} {"custom_id": "Mymensingh_10260_25137_10773_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DLsyZkKtVEyYQuoAs33xxh"}]}]}} {"custom_id": "Faridpur_25650_13851_26163_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Jb7xJGZXKveJrtquwK2uLn"}]}]}} {"custom_id": "Noakhali_14364_8208_14877_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EsG1o144RVGcEiGRgWdQ1q"}]}]}} {"custom_id": "Bhola_17955_34884_18468_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PYTxjuVuQgM6hLgKC4GHbi"}]}]}} {"custom_id": "Sherpur_7182_2565_7695_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3cqMpePenp6mdAxuLDkHd8"}]}]}} {"custom_id": "Rangamati_12312_66177_12825_66690.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HkAdF41ocPwCcmgCdbcnLD"}]}]}} {"custom_id": "Dinajpur_1026_10260_1539_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WUbPp6We1GTXBRMiHwTAmB"}]}]}} {"custom_id": "Jhalokati_13338_6669_13851_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QZx3NnD2ia424teJTeWprV"}]}]}} {"custom_id": "Bogra_19494_28215_20007_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2ofjdtuig1ZJ946pycCb3a"}]}]}} {"custom_id": "Thakurgaon_8721_2565_9234_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hs4NXerM9H4dvHyWBLj9GZ"}]}]}} {"custom_id": "Noakhali_7182_39501_7695_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W5vtoWxvbWhmpNjEy9K9SU"}]}]}} {"custom_id": "Nawabganj_22572_11286_23085_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LY3FyGKXSzXRoEzRSx69X8"}]}]}} {"custom_id": "Sunamganj_27702_19494_28215_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Kub8WwY7xL52qQU6mpN87h"}]}]}} {"custom_id": "Naogaon_7182_5643_7695_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9cB5PSp4droaKmR89zdPdW"}]}]}} {"custom_id": "Jamalpur_10773_9747_11286_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XfuVyunH6ZT2zYZwg9Krjr"}]}]}} {"custom_id": "Sunamganj_35910_4104_36423_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SUormJzFMZ3T6DU1V9UusL"}]}]}} {"custom_id": "Comilla_21546_4104_22059_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-C5TuiXsq3ALEbpTVvVdLGN"}]}]}} {"custom_id": "Rangamati_15903_56943_16416_57456.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UPPHP5xGRUtvh8bGsqvz5L"}]}]}} {"custom_id": "Dinajpur_26676_23598_27189_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1fMXYVTcSa3FhvGhzUyJtJ"}]}]}} {"custom_id": "Nawabganj_17955_11799_18468_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8jQe4QCnta6uJGD4VwTtDe"}]}]}} {"custom_id": "Naogaon_17955_22572_18468_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2kSw6xSavTTmcGTqsy8KdP"}]}]}} {"custom_id": "Shariatpur_9234_11799_9747_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mi15RRFbFePxujPF1D6idR"}]}]}} {"custom_id": "Khagrachhari_18468_24624_18981_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4UugocL89dK5jZdSvzvAJJ"}]}]}} {"custom_id": "Brahamanbaria_4104_21546_4617_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EDRAquPWAFhiME88DDoEQU"}]}]}} {"custom_id": "Jhenaidah_17442_19494_17955_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LpRsAsZZVUaZN2nuuHCnBQ"}]}]}} {"custom_id": "Chittagong_22572_9747_23085_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R3GDoCBwoHkbmGxfwgpfZR"}]}]}} {"custom_id": "Jhalokati_513_18981_1026_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4NbYPhujWDbRhF69nipJyg"}]}]}} {"custom_id": "Rangpur_19494_21546_20007_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GauVAJb2DtfoP5Fmr8a4A5"}]}]}} {"custom_id": "Bogra_30780_21033_31293_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WapbjZ6xNubh8SjHiH6tsZ"}]}]}} {"custom_id": "Satkhira_6669_30780_7182_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SciR9NsvvjHtTWA3ZVx1SY"}]}]}} {"custom_id": "Rangamati_28215_67203_28728_67716.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-As7jedndqeyrdQSrVby4Ze"}]}]}} {"custom_id": "Bandarban_9747_44631_10260_45144.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FbXzJowu6NpE8efF4p79j2"}]}]}} {"custom_id": "Naogaon_7695_1539_8208_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DygNqFnVNkYFftaKDWKJke"}]}]}} {"custom_id": "Bogra_11286_14364_11799_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BnA6ZBPPkxaW6SsiwcaFkY"}]}]}} {"custom_id": "Shariatpur_10260_3591_10773_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mb5jY3ZtUHKDFtjLAGyHDA"}]}]}} {"custom_id": "CoxsBazar_8208_15903_8721_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WnFKr9GDyCkD2yoXCA7iJU"}]}]}} {"custom_id": "Satkhira_10773_38475_11286_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EEWReTcBKbjuVdU5cGFQnA"}]}]}} {"custom_id": "Meherpur_2565_11286_3078_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3AmorBAHyWfsuM46d3QoEY"}]}]}} {"custom_id": "Bogra_26676_4617_27189_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S6cMvg3A3KujRey3Zzqgdy"}]}]}} {"custom_id": "Lalmonirhat_26163_23598_26676_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UNtEw1TkQHUw2pKLUJ7ZHw"}]}]}} {"custom_id": "Sylhet_13338_18468_13851_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YT9Z3iRr7jcKqUTn79dGau"}]}]}} {"custom_id": "Jhalokati_4104_12312_4617_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JQFwwqNXNpoV4F8U675Efe"}]}]}} {"custom_id": "Patuakhali_3078_31293_3591_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XHDhMML8rDG1V3GBH91oN7"}]}]}} {"custom_id": "Kurigram_19494_34371_20007_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HqkrFNcXm6dznYyNaWonZi"}]}]}} {"custom_id": "Jamalpur_10260_12825_10773_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3mEeaKdW3yKzFJ8PS5rfR6"}]}]}} {"custom_id": "Comilla_24111_25137_24624_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8PSpSN7AmruMAcKozH1Cf6"}]}]}} {"custom_id": "Manikganj_5130_3591_5643_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Wwgy5Kvfp5AarDTA3wLfkq"}]}]}} {"custom_id": "Rangamati_17442_66177_17955_66690.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QbqbQAGbrzhKUy8NpEtoax"}]}]}} {"custom_id": "Jamalpur_8208_7182_8721_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KETrPGs6Cm9k7y6io7ch9h"}]}]}} {"custom_id": "Bandarban_19494_7182_20007_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FUogaSWeSeTefg26yjN3cR"}]}]}} {"custom_id": "Bandarban_6156_50787_6669_51300.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-H1XLq5pbRiD2Ma6kRhELUV"}]}]}} {"custom_id": "Sunamganj_5643_4104_6156_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7eS6gPMQiH7dsr2MbmpLtb"}]}]}} {"custom_id": "Chittagong_9747_30267_10260_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NuTUNFB5yWTLZiSNraZANC"}]}]}} {"custom_id": "Chandpur_12312_7182_12825_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RDZver2vMyu3hQvxWiLXmU"}]}]}} {"custom_id": "Chittagong_25650_26676_26163_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HgQD2psu2u5aL7S8Y52B8D"}]}]}} {"custom_id": "Mymensingh_15903_2052_16416_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4UBUQ8a2KVKTtkdRC94SwC"}]}]}} {"custom_id": "Rajshahi_26163_3078_26676_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-174qzzdmsSSzLEKCiKZF4R"}]}]}} {"custom_id": "Bandarban_10260_38988_10773_39501.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-28TCTAyify3XZVDaB1M84d"}]}]}} {"custom_id": "Pabna_23598_24111_24111_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5AokRrPoTT8yQNfvPnJonV"}]}]}} {"custom_id": "Magura_10773_8721_11286_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bw7KmJzzVn33nWZAetr36c"}]}]}} {"custom_id": "Pabna_19494_6669_20007_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EwDhJgjeRiN235qEXYuCVM"}]}]}} {"custom_id": "Thakurgaon_12312_4104_12825_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bb92oKey72zKhZ1JbZqLcE"}]}]}} {"custom_id": "Jhalokati_14877_10260_15390_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RCDPQWA5Fjh5f8DPA9gHY1"}]}]}} {"custom_id": "Nilphamari_8208_11799_8721_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MJ3uG4rYNY8RQyoLoDoro5"}]}]}} {"custom_id": "Feni_8208_2052_8721_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Sdw716j3KHFQoSaCrSjewL"}]}]}} {"custom_id": "Sylhet_18468_1539_18981_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LdkwmwUdp8PkJvdp15nrfG"}]}]}} {"custom_id": "Gaibandha_15390_27189_15903_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-262Da9suSUSgtFD3WbaET4"}]}]}} {"custom_id": "Rangpur_18981_23598_19494_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5AHqMHCh13saHwRx8ZmTH6"}]}]}} {"custom_id": "Sirajganj_4104_20520_4617_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1Hgk9eWCUo7kz3vZurERKT"}]}]}} {"custom_id": "Bandarban_7695_43605_8208_44118.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AgEzh6J8n3UvHYZw351cLY"}]}]}} {"custom_id": "Barisal_17442_8208_17955_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Uishs8KwEAkPGkaqfkPirK"}]}]}} {"custom_id": "Jamalpur_8721_13338_9234_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J2PJ7TqvHAj25z46X2Bqbd"}]}]}} {"custom_id": "Khulna_5130_15903_5643_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R1nAzKTmbPHciH1x9r7MxV"}]}]}} {"custom_id": "Sherpur_14364_6156_14877_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5UuztDANYKfDizJhVHu1Yu"}]}]}} {"custom_id": "Barisal_7182_12825_7695_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Rbiw6B3peRaxZaRgtFx7jx"}]}]}} {"custom_id": "Tangail_14877_35910_15390_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EmmwGPm2XfDeZqCxufESH5"}]}]}} {"custom_id": "Barisal_28728_11799_29241_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Wme6cZ8jZ92okFFRFFNe2b"}]}]}} {"custom_id": "Bogra_30267_18981_30780_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-36NzHzB5kFDqj3CRywmaxP"}]}]}} {"custom_id": "Noakhali_10260_11799_10773_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1eZpZm911c3EyjaVH9NEmV"}]}]}} {"custom_id": "Habiganj_19494_7182_20007_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J2uzxLCdD3K7LsApYkuRxo"}]}]}} {"custom_id": "Bagerhat_11286_46170_11799_46683.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1QhkswQ7HpswrHcoLzzP8N"}]}]}} {"custom_id": "Bagerhat_16929_11799_17442_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8Ne6yLrMbDvqtA44MBhVqW"}]}]}} {"custom_id": "Maulvibazar_10260_27189_10773_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VraottT8KBD5dZCKzbA1wb"}]}]}} {"custom_id": "Panchagarh_11799_14364_12312_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KMNcci5WidwwHFFtX77Ez6"}]}]}} {"custom_id": "Sunamganj_31806_20520_32319_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6VU63oiMxVh868tZmZqcV9"}]}]}} {"custom_id": "Sirajganj_16416_21033_16929_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FBMYZLHsyrHGhp2SPcbaFg"}]}]}} {"custom_id": "Tangail_16929_22572_17442_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GBUH37YyUKULqLqcfNU5oY"}]}]}} {"custom_id": "Noakhali_11286_37962_11799_38475.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XPGuKeR345PRwp75MqKYMW"}]}]}} {"custom_id": "Bandarban_14364_14364_14877_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3gGbgoUWA1TxE7tbB9R4Hk"}]}]}} {"custom_id": "Magura_10773_7695_11286_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MXAS14zkqs28xnECTewYrL"}]}]}} {"custom_id": "Chittagong_35910_32319_36423_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YJdHn29SWi141jm62UsB5g"}]}]}} {"custom_id": "Satkhira_9747_44118_10260_44631.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LyJsJvLdxQLMYbK2gP4741"}]}]}} {"custom_id": "Sunamganj_7182_13851_7695_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-95zwGj7dKfS44kcrW6RL6F"}]}]}} {"custom_id": "Jessore_18981_14877_19494_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-B1eg5LtpktqwZsYPS8bxrQ"}]}]}} {"custom_id": "Sirajganj_14364_20007_14877_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YS7EvbQwmC9hebHfv6Zvh9"}]}]}} {"custom_id": "Comilla_14877_6156_15390_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VT6KHDCTHH3HRq7BuqcBHz"}]}]}} {"custom_id": "Comilla_13338_11799_13851_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8fkaqeHh2sHivsbVm8oyTL"}]}]}} {"custom_id": "Khagrachhari_10260_20520_10773_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-43PEMz6brtdAD8bhBGb9BL"}]}]}} {"custom_id": "Bagerhat_6156_24111_6669_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SaHBRuaoBuUo127oJCJUsU"}]}]}} {"custom_id": "Tangail_5130_23085_5643_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U8voD8a2rvPoGCPRDUwKCR"}]}]}} {"custom_id": "Comilla_14364_5130_14877_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4qQuYu7djfpSaNyygriTH5"}]}]}} {"custom_id": "Barguna_6156_14364_6669_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2tTo54kQDKwdW5Rscthj1u"}]}]}} {"custom_id": "Rangamati_10260_37449_10773_37962.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1JJAyoHqxu75sSxw39ovrn"}]}]}} {"custom_id": "Lalmonirhat_15390_21033_15903_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JvcE3DZ4SPUCddhAu9TWYs"}]}]}} {"custom_id": "Gaibandha_18981_14364_19494_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BcsJsrsDCerEaFxecVHpyM"}]}]}} {"custom_id": "Khulna_4617_61047_5130_61560.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T8mrKSJH3Ru2jX1fZFQyoV"}]}]}} {"custom_id": "Noakhali_13338_3591_13851_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VYtPMPdX797fUV5FsYLg3J"}]}]}} {"custom_id": "Feni_8721_13338_9234_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LTkPrGBnY5fcJ82pnxCXgc"}]}]}} {"custom_id": "Rajbari_13338_10773_13851_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CLCmZtzgwb2sxoTxPz6NN4"}]}]}} {"custom_id": "Sunamganj_3078_9234_3591_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SCd1mXQ44rDVCjfoDHK5MR"}]}]}} {"custom_id": "Tangail_6669_23085_7182_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X5mhSdLT7TpBHUzVAwNBvt"}]}]}} {"custom_id": "CoxsBazar_10260_17955_10773_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TGH8RpFmpnq2DZAtTLdeWZ"}]}]}} {"custom_id": "Rangamati_20520_52326_21033_52839.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Qx68CGx8cvmFUYZ1raWSKJ"}]}]}} {"custom_id": "Narsingdi_6156_9234_6669_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4U8H3KfKDrxBPqkdWyS3TX"}]}]}} {"custom_id": "Kishoreganj_23598_17442_24111_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MXpnBruxzmFpGk66qfrnt6"}]}]}} {"custom_id": "Manikganj_14364_14877_14877_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3QZxK3xBjTERPxNG5Q8XmF"}]}]}} {"custom_id": "Khulna_6156_27189_6669_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-C6nSmeKdvS8CM2y1BpT7iP"}]}]}} {"custom_id": "Dinajpur_14364_15903_14877_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QweXZT3YtnPHhR2ssEL2hV"}]}]}} {"custom_id": "Jhenaidah_15903_14364_16416_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FKN4BiPV6YDtrzE7zmwUsp"}]}]}} {"custom_id": "Sylhet_22059_5643_22572_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YJgmQXa4ruQkPum3JTqdEJ"}]}]}} {"custom_id": "Pabna_7182_16416_7695_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gty1s1Fw5M552VNinioKVy"}]}]}} {"custom_id": "Naogaon_11799_8721_12312_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D1SBQAjVdoFMw5fE12SJFK"}]}]}} {"custom_id": "Sunamganj_16929_12825_17442_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FSC94NTs4T3ATAERNXT9R9"}]}]}} {"custom_id": "Kurigram_6156_1539_6669_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-C5RQe65N5pAH2myq6v7eaS"}]}]}} {"custom_id": "Kushtia_22059_13338_22572_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J46bXHR9dNLPcsDgC9wkz6"}]}]}} {"custom_id": "Bandarban_8208_46683_8721_47196.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UmvCWmyKhEVhrFT4A7wYvD"}]}]}} {"custom_id": "Maulvibazar_5130_30267_5643_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6q3KfYnmD2ApYn28EjkemC"}]}]}} {"custom_id": "Bagerhat_16929_17955_17442_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L5uepMMy7p1hQRv8dXMnsX"}]}]}} {"custom_id": "Barguna_2565_22059_3078_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7Jxa8QUGi4VQFbFYD1eiCc"}]}]}} {"custom_id": "Netrakona_13338_27702_13851_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-56yGEN4T6vs4j5r66N33fA"}]}]}} {"custom_id": "Chittagong_6669_11799_7182_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4CPPcQsuMYxPCagjhdRaB3"}]}]}} {"custom_id": "Brahamanbaria_19494_25650_20007_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4p3F5L8rZmzFiHfSqgPoQK"}]}]}} {"custom_id": "Bagerhat_14364_34371_14877_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W2pXWAXVAjCPyjFtA5hYGD"}]}]}} {"custom_id": "Feni_3591_22572_4104_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Udza63XPNaQLDyh99mB1jA"}]}]}} {"custom_id": "Tangail_16416_22059_16929_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4pHLpoLoWtyTMPut4KBSan"}]}]}} {"custom_id": "Pirojpur_0_8721_513_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-G6XCsQkb7HT11nUEqZSTVi"}]}]}} {"custom_id": "Chittagong_13338_16416_13851_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9QML2tDuwWRBb67U682EZR"}]}]}} {"custom_id": "Rangamati_9747_35397_10260_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8mC8oGtabkkitWdjJEpnsd"}]}]}} {"custom_id": "Magura_14364_18981_14877_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SBzMSZbST8RdgMACZAKPaU"}]}]}} {"custom_id": "Kurigram_17442_30267_17955_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GKj1L9NRr6Ujdk3AytTvHZ"}]}]}} {"custom_id": "Patuakhali_9234_8208_9747_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5aj4pAKZow5P5us9tzRRwv"}]}]}} {"custom_id": "Habiganj_16416_16929_16929_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GJPuPps4kcdNbcn2YeH6Vk"}]}]}} {"custom_id": "Magura_6156_15903_6669_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8AVxAUrKuWKri6JqmDzHpJ"}]}]}} {"custom_id": "Bandarban_7182_48222_7695_48735.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4okrg4V1Hwsh1zoRqR41Bo"}]}]}} {"custom_id": "Kishoreganj_21546_9234_22059_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2N3VQD2YkpgGHKcsjdYXdY"}]}]}} {"custom_id": "Joypurhat_14364_12312_14877_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9MCai55oQyjqxpaoSzw5A9"}]}]}} {"custom_id": "Tangail_6156_8721_6669_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4Z9bwfxpeTspLL2LBs3VMy"}]}]}} {"custom_id": "Habiganj_18468_1026_18981_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CAVMiZJW1nRRjXFnCXnrZg"}]}]}} {"custom_id": "Barisal_26163_12312_26676_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PzHrBfC7qRHZ1LAXxfPFqR"}]}]}} {"custom_id": "Khagrachhari_12312_26676_12825_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ws9aFChucdEbWwftLAKaKk"}]}]}} {"custom_id": "Natore_5643_11799_6156_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1W3sgMxZgatCR1koKBnRip"}]}]}} {"custom_id": "Bagerhat_13851_47709_14364_48222.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QAQToFknDswUcn7jSqFWpv"}]}]}} {"custom_id": "Jamalpur_4104_30267_4617_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Q5Hfq886Kx3ZKsaWacaVFx"}]}]}} {"custom_id": "Kishoreganj_15903_18981_16416_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DjifQrGLJ34NXvVvr9Pwkr"}]}]}} {"custom_id": "Tangail_513_13338_1026_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CUZuD3ADxdpFVSQXCrirk5"}]}]}} {"custom_id": "Bogra_11286_18981_11799_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DHY7RVFQ71QCbE8Kxfx8zY"}]}]}} {"custom_id": "Chittagong_38988_38475_39501_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2jAmxHRQtf46Fwvj3DZrBq"}]}]}} {"custom_id": "Feni_513_14364_1026_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5kwPJfdbgHA1Ki1mVNHJhx"}]}]}} {"custom_id": "Dinajpur_3078_21033_3591_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KFAXNhXo8612zKgZKCipVb"}]}]}} {"custom_id": "Kishoreganj_30267_16416_30780_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L7PySmJ7JMCe2uKfH9qc3H"}]}]}} {"custom_id": "Kurigram_19494_15903_20007_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2T6L8dxr2zxqxh1aPXJemn"}]}]}} {"custom_id": "Naogaon_30780_33345_31293_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Xz5H1uTCWDM9J5zu3biUGB"}]}]}} {"custom_id": "Sunamganj_25650_17442_26163_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AK5pcqWJatKwjRKqQDutRi"}]}]}} {"custom_id": "Noakhali_7695_23085_8208_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hbug7vVHQt9qk7zddR8pzx"}]}]}} {"custom_id": "Gaibandha_17955_14877_18468_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-My2t8r5pCNtP2wRikeQj8x"}]}]}} {"custom_id": "Patuakhali_20520_18468_21033_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EYHvhgRrAMuNFjUmSPFQjz"}]}]}} {"custom_id": "Bagerhat_7695_39501_8208_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5aVurDXeksycdm889BwA7D"}]}]}} {"custom_id": "Rangamati_20520_33345_21033_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9GKyzmHbNKwZ4rdq4tbbhn"}]}]}} {"custom_id": "Gopalganj_20520_12312_21033_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QQsVgUob5CtxDatUuWFyWM"}]}]}} {"custom_id": "Netrakona_26163_23598_26676_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6M7VMUzGbmVj1qocmS9VFo"}]}]}} {"custom_id": "Bagerhat_18468_11286_18981_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-G596LAa5TgKAW3hZRYWx1o"}]}]}} {"custom_id": "Patuakhali_8208_36423_8721_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-696fpCcciQ9iQjcwjmphL9"}]}]}} {"custom_id": "Narayanganj_8721_7182_9234_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BByyP9xLWYSnudiTmj91mA"}]}]}} {"custom_id": "Barisal_9747_3078_10260_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6B2s1XE97G3gGwamBP8551"}]}]}} {"custom_id": "Gaibandha_9747_26163_10260_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U1AKUpnf7ULnHJNKd7yL1z"}]}]}} {"custom_id": "Narsingdi_12312_13851_12825_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1NGCqZY14Hey1ezgYuEWBH"}]}]}} {"custom_id": "Rangamati_15903_15390_16416_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RkbgqJW2xGbqU6f6sDJsWT"}]}]}} {"custom_id": "Barisal_27189_8721_27702_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PPPt5pqiCXKNESFKbxxWmm"}]}]}} {"custom_id": "Bandarban_4617_25650_5130_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WE3uiiPKfRzM9YCTUPwB3o"}]}]}} {"custom_id": "Jhalokati_13338_8721_13851_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8QdvEtgGjt2tVuAUZxd6q1"}]}]}} {"custom_id": "Sylhet_18468_5643_18981_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YPkAJeFH53r4m9KsL8yEFi"}]}]}} {"custom_id": "Bandarban_24624_53865_25137_54378.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LhJazfRYMA696gzNmF4kRa"}]}]}} {"custom_id": "Dinajpur_11799_9234_12312_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6mCQU5y4VbLJdb2ZZwMcWy"}]}]}} {"custom_id": "Netrakona_11799_25137_12312_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GNhkEBoPeGikN72QqGnUFQ"}]}]}} {"custom_id": "Brahamanbaria_16416_23085_16929_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NdMXneE3F2scrgfJRzedMp"}]}]}} {"custom_id": "Comilla_3078_8208_3591_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DP7oyGGb8iZSYKhKqThudQ"}]}]}} {"custom_id": "Rangpur_8721_9747_9234_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ledkn2E3nwF5UGsf6pE3ZF"}]}]}} {"custom_id": "Chittagong_15390_16929_15903_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-C8BXoX5u4Us6LKwciDnymW"}]}]}} {"custom_id": "Khulna_9234_48735_9747_49248.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Rfpu1SpuzKFxcYi2fGADrH"}]}]}} {"custom_id": "Bogra_22572_26163_23085_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-G2gejfW8y2iwdPD3PnnsiJ"}]}]}} {"custom_id": "Munshiganj_11286_1539_11799_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K9aQetqmhoVQumnmEcGKY9"}]}]}} {"custom_id": "Joypurhat_6669_5643_7182_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-G4YwHeCcJ2qBS4BkLEHnAA"}]}]}} {"custom_id": "Nawabganj_17955_7695_18468_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HeXkQzFevTrZK9oyYDdQWH"}]}]}} {"custom_id": "Barisal_10773_9747_11286_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3i9g5UAUM3wJjBJeH9ZXXs"}]}]}} {"custom_id": "Sirajganj_15903_22572_16416_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FFxgRqaciRWWEzaDmm8QLK"}]}]}} {"custom_id": "Habiganj_21033_10773_21546_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CwPLaHtBwvPR61W1mb4Z5e"}]}]}} {"custom_id": "Mymensingh_8208_33858_8721_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NQM1XpkUKq2TTFo8EehPiX"}]}]}} {"custom_id": "Kushtia_23085_22572_23598_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TQ3Ux7uBfwQrjBQuCuS5NV"}]}]}} {"custom_id": "Maulvibazar_25137_12825_25650_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RGcWJgpcUJ6AzRDSMf4wob"}]}]}} {"custom_id": "Patuakhali_16929_3591_17442_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GazCg2FUKJUR7j3N8K3Euu"}]}]}} {"custom_id": "Mymensingh_10773_41553_11286_42066.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CsvFeUGyQByTGLpaMJBCTi"}]}]}} {"custom_id": "Gazipur_13338_6669_13851_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LmbgRDmdagJ4rfikeiA58q"}]}]}} {"custom_id": "Maulvibazar_13851_21546_14364_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FuaCCZvbqmGxnGXZS9GGEZ"}]}]}} {"custom_id": "Brahamanbaria_15903_24624_16416_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FB1C2PuNAnoMBhbrwwNHFt"}]}]}} {"custom_id": "Rangpur_14364_17955_14877_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ccq8x4D1P7n2wkHVSniKJy"}]}]}} {"custom_id": "Noakhali_17442_13851_17955_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R2uhHryUnKDGviACPZsno9"}]}]}} {"custom_id": "Sylhet_2565_29241_3078_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9rRPU3kgWfmsUJWyxWLoLk"}]}]}} {"custom_id": "Chittagong_34884_53865_35397_54378.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AzdDdf3a1PCrZ9jZ4ye4Xo"}]}]}} {"custom_id": "Bandarban_23598_44631_24111_45144.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AJtgg4phdbAnfroCDjp1sN"}]}]}} {"custom_id": "Brahamanbaria_14877_25650_15390_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Dw1YF3BrRCGcHzWb9kLhtr"}]}]}} {"custom_id": "Jamalpur_5643_28728_6156_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D1UCDaKEDBepHiTo1dYaLk"}]}]}} {"custom_id": "Chandpur_19494_8721_20007_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6dUqzJD2xq6ivSinLoLuqN"}]}]}} {"custom_id": "Bagerhat_0_56943_513_57456.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-H5wcWJHma7yzvrWvNU9JRZ"}]}]}} {"custom_id": "Jessore_16416_17955_16929_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1TAtzyW7p9TL2DXjJg7h8a"}]}]}} {"custom_id": "Khagrachhari_11799_28215_12312_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WAqXEUB4Jr3w7US7Z4md8B"}]}]}} {"custom_id": "Mymensingh_19494_1539_20007_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DEj5J2Jq32WazfWCbrGfB3"}]}]}} {"custom_id": "Kishoreganj_23085_8721_23598_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AnAcvWPG9UAqEZHpg5bNXU"}]}]}} {"custom_id": "Bhola_11799_37962_12312_38475.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BxmNy4Sx4sejTYKS6jg1Tc"}]}]}} {"custom_id": "Bandarban_18981_36423_19494_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XQ3EWHoLCBAM3Fd4a46Gzr"}]}]}} {"custom_id": "Mymensingh_10773_32319_11286_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HkuSUJw3U8ogHKDaC51kSo"}]}]}} {"custom_id": "Khulna_11799_30267_12312_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9vpZWkAqa22oLDQKRfNeUc"}]}]}} {"custom_id": "Faridpur_26163_16416_26676_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KLaerSkseLa7jnNuEm55A7"}]}]}} {"custom_id": "Bogra_11286_16929_11799_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KzZtpBYr2KUNv3dUErrPw8"}]}]}} {"custom_id": "CoxsBazar_10773_35397_11286_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4Kk1JR6szWr8SLa5d3ErJP"}]}]}} {"custom_id": "Natore_13338_19494_13851_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F3KLPuzBZB7gkJwwXdKUiG"}]}]}} {"custom_id": "Chittagong_25137_40014_25650_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HcNXqvgoWDi3EHLKFxCepW"}]}]}} {"custom_id": "Madaripur_2052_14877_2565_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KK95HkWnqLuM1JvNBy3tfN"}]}]}} {"custom_id": "Maulvibazar_28215_9747_28728_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LKC8CVhBbWoZJMX9SghKn2"}]}]}} {"custom_id": "Jamalpur_8208_18468_8721_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GSPcUj13BWjCWTXLPvsnpK"}]}]}} {"custom_id": "Madaripur_11286_6669_11799_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9Qo9LEQeKn2sC7zQWGAbFn"}]}]}} {"custom_id": "Jamalpur_6156_8721_6669_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XGLywmsG5BJpZ5d3RE8mpe"}]}]}} {"custom_id": "Habiganj_20520_18981_21033_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VRw4NMff9gDneDvtjJBqUS"}]}]}} {"custom_id": "Jhalokati_5643_6156_6156_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LYbuRVmN2rQKbE3KT2HkwA"}]}]}} {"custom_id": "Barisal_12312_17955_12825_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VUu3qBm72ncZd6gZSpWR7s"}]}]}} {"custom_id": "Dinajpur_18981_27702_19494_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LQG8T1wT12AGkjD7somcj3"}]}]}} {"custom_id": "Khulna_12312_53352_12825_53865.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9NV5nMS4sSyjzvCYvdUa8h"}]}]}} {"custom_id": "Rangpur_15390_9234_15903_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RH5GTPVUzdrutmazbpjavA"}]}]}} {"custom_id": "Satkhira_15390_58482_15903_58995.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-84gQ2DTxsMfJkW4ekZUqUb"}]}]}} {"custom_id": "Shariatpur_1539_11799_2052_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3ZKVbaCutffrWeUvdfQPNP"}]}]}} {"custom_id": "Barisal_24111_5643_24624_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TND8L7axi9HS4bU5KD5kUV"}]}]}} {"custom_id": "Naogaon_11286_4104_11799_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7DiPms8smWemdEQ4Etjsu9"}]}]}} {"custom_id": "Sylhet_1539_20007_2052_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fxs3UNgnC2wgzEY3sMmrhS"}]}]}} {"custom_id": "Satkhira_3591_10260_4104_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-64m2bR99wrDKQ1LJciKjBQ"}]}]}} {"custom_id": "Rangamati_12312_27189_12825_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NFsY5xaGTMf2mhojYVuFnS"}]}]}} {"custom_id": "Jamalpur_17442_26676_17955_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FCx8Lpy44p7kqfpAXW88aG"}]}]}} {"custom_id": "Nawabganj_12825_23085_13338_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ArEVAPRkoXs224h6Feksff"}]}]}} {"custom_id": "Naogaon_18468_11286_18981_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DXNBkK1sSkb7YFkoijBFQ3"}]}]}} {"custom_id": "Rangamati_20520_23085_21033_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1L3J5hfk8Dm9aWcWnkZ2kQ"}]}]}} {"custom_id": "Sylhet_33345_14877_33858_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1LrdHtxoR1eqC5gRV73okr"}]}]}} {"custom_id": "Comilla_17442_32832_17955_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UmjhxX9rke4t7TCsHnpgGA"}]}]}} {"custom_id": "Kishoreganj_29754_6156_30267_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6kkdKe2p8xz6Q69PVBJLEo"}]}]}} {"custom_id": "Barisal_21033_25650_21546_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PxHSceVHy1u5pSWUetyvth"}]}]}} {"custom_id": "Narsingdi_17442_16929_17955_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3uGUzsNA1nqsZJbR8rfRUt"}]}]}} {"custom_id": "Kurigram_16929_16929_17442_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KJzwP1L772tiKFijk4mx2u"}]}]}} {"custom_id": "Chuadanga_8208_15390_8721_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HzE1q77shunjQUb4rMzJ21"}]}]}} {"custom_id": "Bagerhat_7695_50787_8208_51300.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EuB77RVhjgYi9L5CionzG9"}]}]}} {"custom_id": "Dinajpur_13851_13851_14364_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CM3uXb3k5EoyDqpFz9QBCR"}]}]}} {"custom_id": "Joypurhat_3591_10260_4104_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-75rJ8CLUmDTPeym4NsEqUd"}]}]}} {"custom_id": "Dinajpur_26163_27189_26676_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T5sj2Bc2zH941V3gpNinL2"}]}]}} {"custom_id": "Lakshmipur_5643_9747_6156_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4WK68WGiy2Qy4TiqYq5KEk"}]}]}} {"custom_id": "Rangpur_19494_24624_20007_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XFVByq55LSPmj7GyDv1fZr"}]}]}} {"custom_id": "Sirajganj_16929_6669_17442_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AnZTgg23czxPamM7cMcxb4"}]}]}} {"custom_id": "Madaripur_17955_22059_18468_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Je9MS2fmG1LmyK6qW6bgv1"}]}]}} {"custom_id": "Kurigram_5643_23085_6156_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-McheknfPsgUm3QvVvqtph5"}]}]}} {"custom_id": "Jhalokati_2052_11799_2565_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WN4tWeDC2QRgahuy4YX76q"}]}]}} {"custom_id": "Feni_5643_24111_6156_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Eoap5GAUkFTPo1v7T1oSkr"}]}]}} {"custom_id": "Mymensingh_22572_16416_23085_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1ZjpY7XGBAybAEQ6PSq7fW"}]}]}} {"custom_id": "Bogra_7695_16416_8208_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LAjprXxiKr81HBvALcVx6z"}]}]}} {"custom_id": "Panchagarh_2565_6669_3078_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Sybr2pxM5L5oY684jJvcvU"}]}]}} {"custom_id": "Kishoreganj_21546_14877_22059_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D3DLMTGuJdw8GsxanyA4kF"}]}]}} {"custom_id": "Feni_12825_13851_13338_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3hm1CbfwANyDza3zHHjhVD"}]}]}} {"custom_id": "Naogaon_7182_18468_7695_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ene1AFDTaVC24tW2GB3sKW"}]}]}} {"custom_id": "Narayanganj_1539_10260_2052_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BAjoekvsYXCTcXcFojtC4v"}]}]}} {"custom_id": "Gazipur_513_16416_1026_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vc3MCXnfiBRPnKvnG5f7fh"}]}]}} {"custom_id": "Satkhira_12312_15903_12825_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Q174dKHAoS7wKEbBPxXAFv"}]}]}} {"custom_id": "Tangail_15390_24111_15903_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3qHGgpq8BZjJKktNJbLFWT"}]}]}} {"custom_id": "Chittagong_23598_9747_24111_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U1kVx35CH3zeM6SwVpq1wR"}]}]}} {"custom_id": "Manikganj_5643_6669_6156_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1CTSchQCoUqLyj3YbQ8EK3"}]}]}} {"custom_id": "Sunamganj_30267_28728_30780_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-33vVegNqWonGp9bjqhx5F9"}]}]}} {"custom_id": "CoxsBazar_20007_45657_20520_46170.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TmmpCUKdsLFKSS4t9yBvLr"}]}]}} {"custom_id": "CoxsBazar_24111_58995_24377_59508.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YV4zYGKtrp6AZ1RMgdgCS4"}]}]}} {"custom_id": "Mymensingh_10773_18981_11286_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JiZ1exeJ7dmfZFNFGYMRfC"}]}]}} {"custom_id": "Maulvibazar_15390_29241_15903_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GPiQeRux8Q4TStpFFXoNvX"}]}]}} {"custom_id": "Sirajganj_8721_12825_9234_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7nsr9eh2W6a615WBjn6HZW"}]}]}} {"custom_id": "Barguna_12825_23598_13338_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Py2LWW7RqiTr42QhJeo5Ry"}]}]}} {"custom_id": "Lakshmipur_8721_20520_9234_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-H7YXBH2chqWyQuvo2RiffN"}]}]}} {"custom_id": "Munshiganj_3591_3078_4104_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CWK6E5YS8UDQ4EnDnTitrQ"}]}]}} {"custom_id": "Noakhali_17442_12825_17955_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PqxaPXp87HXtHmgRQiNFEY"}]}]}} {"custom_id": "Mymensingh_14364_18468_14877_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BSqjCTvp5yzr1GSiHK8atp"}]}]}} {"custom_id": "Bagerhat_5130_53352_5643_53865.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8QLLjxRDz8w2mu9c897cLz"}]}]}} {"custom_id": "Jessore_6156_2052_6669_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8Ss6amdf39AhV251cvqQ9W"}]}]}} {"custom_id": "Gaibandha_7695_29754_8208_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QPZHoKAkKPAaFbGT1RRZyY"}]}]}} {"custom_id": "Bandarban_10773_27702_11286_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M3wMcFftRnp59cjNVejXnR"}]}]}} {"custom_id": "Rangamati_12825_49761_13338_50274.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D3a3B9DrPb1VZhg5PbK8oQ"}]}]}} {"custom_id": "Natore_20520_17442_21033_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Dm573vK9tS3QWPJcwWqZgz"}]}]}} {"custom_id": "Kurigram_3591_18981_4104_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WRrue8ymB5VtWpyPCciHDS"}]}]}} {"custom_id": "Naogaon_12312_12312_12825_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4uhVYPQ1M8U4oKSkdLhWhP"}]}]}} {"custom_id": "Madaripur_8721_10773_9234_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VgP5a34kjkF1RXU2Yti36M"}]}]}} {"custom_id": "CoxsBazar_13851_17955_14364_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DfymkX1xfy54mavizTjFx2"}]}]}} {"custom_id": "Sherpur_6669_4617_7182_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DDSuNeg6cGeVUndBm5Dp9n"}]}]}} {"custom_id": "Satkhira_13338_31293_13851_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2xbBbAFBPe6yCL9Dz962SA"}]}]}} {"custom_id": "Mymensingh_17442_29241_17955_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7j3cJ2rPD8eVcy3CQ7LCP3"}]}]}} {"custom_id": "Sunamganj_29754_5130_30267_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4DRpajAng4wZc7H6Xv3Mqo"}]}]}} {"custom_id": "Chittagong_1026_16929_1539_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5nCYC71zJW2fWEesfW2Q7J"}]}]}} {"custom_id": "Netrakona_17955_13851_18468_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WFPsqYXPY1wTmcw3UY7VwB"}]}]}} {"custom_id": "Meherpur_4617_8721_5130_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F5xisu2RyVi3VjpDNPUfmC"}]}]}} {"custom_id": "Khulna_11286_20007_11799_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R6QUC8QjQs2bvFH5k8RwY1"}]}]}} {"custom_id": "Habiganj_13851_9234_14364_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BkAWZpCG8LGXMKk63uNeBp"}]}]}} {"custom_id": "Barguna_6669_16929_7182_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PsPoG5G1qm5fGc1KHswcZh"}]}]}} {"custom_id": "Pirojpur_4104_8208_4617_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-86oHVDbDNMwnsry8UjZWNN"}]}]}} {"custom_id": "Feni_5130_20520_5643_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DPejBMX4sAAjm3Gr4MJ7ET"}]}]}} {"custom_id": "Rangpur_23598_18468_24111_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mj4rMSA5mXZpdjQKyVSid5"}]}]}} {"custom_id": "Sunamganj_16416_4617_16929_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fcd7psPQj7ydMeKGBnWpiF"}]}]}} {"custom_id": "Gaibandha_17955_30780_18468_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TvXACkQa7vmNqKTTULQ9pd"}]}]}} {"custom_id": "Chuadanga_1026_10260_1539_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1yGjoVRNvn5rrnzHGsbWdM"}]}]}} {"custom_id": "Pirojpur_13338_9234_13851_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GZb5MkfGu6BTiqxnPGbbrG"}]}]}} {"custom_id": "Patuakhali_22572_26163_23085_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-C54gFbGKxfC31F5c2XQLrq"}]}]}} {"custom_id": "Narail_15390_7695_15903_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JRupmRg46997BDEBoVeQz1"}]}]}} {"custom_id": "Tangail_14364_28215_14877_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9u8v3k63Tj2Mk6DxN5Co3o"}]}]}} {"custom_id": "Brahamanbaria_2565_26163_3078_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XaCrespAvW9VJxMxJ1QXkv"}]}]}} {"custom_id": "Noakhali_8721_47709_9234_48222.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BSRdjx9h5AKqtJfuP8qWgx"}]}]}} {"custom_id": "Comilla_7182_6156_7695_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DgtXdeRfSCp5HCWYUCjJDi"}]}]}} {"custom_id": "Brahamanbaria_23085_4617_23598_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Rpd1Tq7PaP84V8ABfuwTPa"}]}]}} {"custom_id": "Faridpur_17442_18468_17955_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WjdR5mDTckN8RwK35F7K3q"}]}]}} {"custom_id": "Bandarban_23085_44118_23598_44631.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CgJ71WJaYHuCa8ncdcG7g6"}]}]}} {"custom_id": "Khulna_13338_6156_13851_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NUXHue6LysNvJty74Ppqpd"}]}]}} {"custom_id": "Comilla_23598_17955_24111_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WZxNWBUAuBY45x98Ef3XyC"}]}]}} {"custom_id": "Dinajpur_22572_30267_23085_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EpwgDpNcg7VUBD9Q9VoTxC"}]}]}} {"custom_id": "Bandarban_12825_31293_13338_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2bjwMPfD9D67TZWecwjCPS"}]}]}} {"custom_id": "Sylhet_31806_5643_32319_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CnvsRzASdu2YLBXwsv2wBc"}]}]}} {"custom_id": "Pabna_23085_12825_23598_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JZeNHzUCZxZ9G2cvaH1ukQ"}]}]}} {"custom_id": "Bogra_25137_26676_25650_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RrDM8gKgcPWjBRi12E1axt"}]}]}} {"custom_id": "Kishoreganj_0_9234_513_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-STQfSLwpEa9NCQg7hkyy4J"}]}]}} {"custom_id": "Bandarban_10260_24111_10773_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Joefw9wGoZh7ae8AYoy9Ra"}]}]}} {"custom_id": "Madaripur_12312_17955_12825_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5S4GwZjscQHgBw5uahzb53"}]}]}} {"custom_id": "Kushtia_17955_11286_18468_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1dEtSZdVSN1ZhTrDnsKKz6"}]}]}} {"custom_id": "Chittagong_11286_4617_11799_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CDnorttxnHDNpPHywrectm"}]}]}} {"custom_id": "Tangail_15390_15903_15903_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ARziGPxnD88iAZ2RZampZm"}]}]}} {"custom_id": "Munshiganj_10773_8721_11286_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bmfj8YvdeMjtcjwCp9TcnQ"}]}]}} {"custom_id": "Rajshahi_17955_4104_18468_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GYfZNTz4s3iHz82eqjy4KU"}]}]}} {"custom_id": "Chandpur_15903_13338_16416_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3viXMXKYdYKMX5dG2kQVxu"}]}]}} {"custom_id": "Comilla_14877_21546_15390_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WB7v4uRMRmhQnCz3jmTRwC"}]}]}} {"custom_id": "Pabna_18981_20520_19494_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K1wohEVZRPGqJ6Q9Q3vrbB"}]}]}} {"custom_id": "Satkhira_9747_28728_10260_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HgAVpzvjJmL11JUPtSa3ZR"}]}]}} {"custom_id": "Barisal_17442_1026_17955_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CNEA9RfEh77ejnNdhDfEbg"}]}]}} {"custom_id": "Nilphamari_7695_3078_8208_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EgLDoz7rFJmV9duF5fNURY"}]}]}} {"custom_id": "Chittagong_30267_20520_30780_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6DobAR6ppfDYwGj2LYcww6"}]}]}} {"custom_id": "Khagrachhari_5130_45144_5643_45657.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ENYKxbUdEuXjeRAczgRQcj"}]}]}} {"custom_id": "Lakshmipur_5643_10260_6156_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2NKJFDrbgHL9JsQP1JTTNh"}]}]}} {"custom_id": "Jessore_10260_513_10773_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3GNQqw3ZTKPcFiFrNkj5Xd"}]}]}} {"custom_id": "Nilphamari_6156_27702_6669_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-99mszRYKh2TqxqPLPTQASd"}]}]}} {"custom_id": "Bhola_5643_14877_6156_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FGaAgfRx6WMgBBzHC67nvj"}]}]}} {"custom_id": "Bhola_17442_15390_17955_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AAKAoM48YuJRpyQdNew66h"}]}]}} {"custom_id": "Bhola_18981_36423_19494_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L9dzrg6z5u4A9JWZTbsqMm"}]}]}} {"custom_id": "Noakhali_2052_16929_2565_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LBXmfFKcXrGt77y7zHFzna"}]}]}} {"custom_id": "Naogaon_4617_3078_5130_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QWC37nL5AS6KttL14jKrMK"}]}]}} {"custom_id": "Rajshahi_11286_10260_11799_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HTaywND7o62Yc8UHj2dHnT"}]}]}} {"custom_id": "Comilla_11799_10773_12312_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R3xiYg8MhDmz4cRNvV8iWD"}]}]}} {"custom_id": "Tangail_14364_21546_14877_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JJtEpKjPW1TveWdAQKStcz"}]}]}} {"custom_id": "Chittagong_18468_14364_18981_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KENoWC9n7XCh9mX3ioPBh3"}]}]}} {"custom_id": "Gazipur_11286_8208_11799_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M7MoWs1HBvAZex1JWc8yUq"}]}]}} {"custom_id": "Barisal_23598_5643_24111_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-C3VqunMYybP8CvtT3EwGar"}]}]}} {"custom_id": "Sylhet_24624_10773_25137_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HUtbQAcrFWkTNjiw7hQivB"}]}]}} {"custom_id": "Patuakhali_21546_8208_22059_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QhoQiAgcS6HjkUfZBECvch"}]}]}} {"custom_id": "Narayanganj_10773_8208_11286_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FM3wnScVMYg8CY16svPaET"}]}]}} {"custom_id": "Bogra_15903_16416_16416_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4qjCKn6CUy7TgewWbE8NGh"}]}]}} {"custom_id": "Barisal_21033_16416_21546_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DgT7YRUBD1vHQAjo9f8d1z"}]}]}} {"custom_id": "Bagerhat_2565_45657_3078_46170.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U9wWft2jcEme4K6pzxV8qs"}]}]}} {"custom_id": "Bagerhat_10773_29754_11286_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UZCJmgrLjPKQR7dbgrd89v"}]}]}} {"custom_id": "Kushtia_25137_11286_25650_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1i4AZnZ9dND1n5Jjv6YEJV"}]}]}} {"custom_id": "Gazipur_11799_5130_12312_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KHtrfiLuP1QM46GKWnL9PC"}]}]}} {"custom_id": "Rangamati_25137_53352_25650_53865.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3suxmLUydNP8QDUPePHMkW"}]}]}} {"custom_id": "Chittagong_24624_41040_25137_41553.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RMwtGNx4Hp4jxeU6yKqvDh"}]}]}} {"custom_id": "Narsingdi_17955_9234_18468_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JMFEy3t39sQnjy6FUg2CXX"}]}]}} {"custom_id": "Chittagong_25650_32832_26163_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UXe287PUN5gwEL4tfY9LyT"}]}]}} {"custom_id": "Netrakona_28728_17442_29241_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6Z54kwZRfszHKZJU3FTiZG"}]}]}} {"custom_id": "Bagerhat_5643_52326_6156_52839.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SW5JSxqdwe2QZtgFk8rFBC"}]}]}} {"custom_id": "Khagrachhari_14877_35397_15390_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RuQQgfaz2uDfCmW9bk7JwT"}]}]}} {"custom_id": "Jhenaidah_23598_17955_24111_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HWnSqQhcDjsNXXh74MVSGf"}]}]}} {"custom_id": "Rangpur_16929_14364_17442_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PkAbwNnmYNanw8yFjgMNND"}]}]}} {"custom_id": "Satkhira_8208_52326_8721_52839.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-B2NuT8vaCSpeNQkKZynP7v"}]}]}} {"custom_id": "Jamalpur_6669_3591_7182_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VbzStvDki32pqqW9ydqD7S"}]}]}} {"custom_id": "Bandarban_9747_33345_10260_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hy7pmDUcDFqViqd6ByXXdM"}]}]}} {"custom_id": "Rangamati_15903_13851_16416_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1FHH74CBPpergwoPu2tnCn"}]}]}} {"custom_id": "Khulna_8208_26163_8721_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KK5fvHP9ktFgmPFM62XvKM"}]}]}} {"custom_id": "Bandarban_17442_31293_17955_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Tu2vCqHCmiRmavdjToPwcR"}]}]}} {"custom_id": "Bhola_16416_23598_16929_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XYwgBeWQjBkVEQe41ijgMK"}]}]}} {"custom_id": "Naogaon_15903_8208_16416_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GcDvfZTo1Zf2mLBhY9QmrL"}]}]}} {"custom_id": "Lalmonirhat_7182_7695_7695_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-22VKMvn5rCAjozjqNMsTUY"}]}]}} {"custom_id": "Maulvibazar_8208_15390_8721_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-USSkCzfMS31eXeKgaxSKfu"}]}]}} {"custom_id": "Barisal_8721_6669_9234_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WEJb1hojXm1seqZ49waQQu"}]}]}} {"custom_id": "Natore_19494_10260_20007_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M2LDUK7YvqQhkbrWLcUybR"}]}]}} {"custom_id": "Mymensingh_9747_38988_10260_39501.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WrKDnpBXixwrT8GJUnHcTd"}]}]}} {"custom_id": "Mymensingh_24624_25137_25137_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DfyyQcBTXhNYhtBF9vz6Nd"}]}]}} {"custom_id": "Comilla_14364_14877_14877_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HzpF83Cj4Jde4jsGKbjgXF"}]}]}} {"custom_id": "Sylhet_15390_3078_15903_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-URVvScGt2cxwUdCkChYovm"}]}]}} {"custom_id": "Sunamganj_18468_13338_18981_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YN7iMQgCScUMsjyjhL4D1g"}]}]}} {"custom_id": "Dinajpur_25137_35910_25650_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ErMUc1NEE3k6WpbEWz42Ri"}]}]}} {"custom_id": "Mymensingh_5643_23085_6156_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WLS25CtokTUnnwC21cDsoo"}]}]}} {"custom_id": "Khagrachhari_7182_21033_7695_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M9tzKrKDXMHnUqXFGvmHcq"}]}]}} {"custom_id": "Rangpur_11286_8721_11799_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7boPjtDHN9joMhgdkkoGNk"}]}]}} {"custom_id": "Bhola_9747_20520_10260_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TUQkAoYaRf1SCFz88eB4Me"}]}]}} {"custom_id": "Brahamanbaria_22059_11799_22572_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UfMsG2EvWJj3cgPXK9NYdx"}]}]}} {"custom_id": "Kurigram_11286_38475_11799_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-STkfYsyQwZEpwU4nvVk8BL"}]}]}} {"custom_id": "Kurigram_5130_17442_5643_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ku9kdAwphnYTJfmzu4KQSz"}]}]}} {"custom_id": "Kishoreganj_10260_6669_10773_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5VHy97LLEefsKJmD2o74Gj"}]}]}} {"custom_id": "Bagerhat_8208_22572_8721_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Qien1bBrqhcAXWa7583jQj"}]}]}} {"custom_id": "Bagerhat_7182_9234_7695_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6ujnchzgfowB1tqJDEJPvg"}]}]}} {"custom_id": "Sunamganj_21546_21033_22059_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UXzfxURXMkUFn6pqLNtDij"}]}]}} {"custom_id": "Khulna_513_9747_1026_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YWf1WZd2V5Q8FoBTtRkBga"}]}]}} {"custom_id": "Barguna_5643_16929_6156_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YJVmJZ6VMqVFed314Xopqq"}]}]}} {"custom_id": "Lalmonirhat_9747_12312_10260_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V9GxchP94qf6SHPekKr6up"}]}]}} {"custom_id": "Satkhira_9747_26163_10260_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EiXExwz5cwUeGmX5xiRAsG"}]}]}} {"custom_id": "Sunamganj_31806_3078_32319_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FqwgqrVGo6UyepgerT8rFQ"}]}]}} {"custom_id": "Natore_10773_9747_11286_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S9yUCyiPELkP1uTeoooSUu"}]}]}} {"custom_id": "Habiganj_2052_10773_2565_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L1epAs9RAAfixK9M45HeS3"}]}]}} {"custom_id": "Netrakona_18468_9747_18981_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1LBs1t8tiRSyEcCH48hT1g"}]}]}} {"custom_id": "Bogra_12312_6669_12825_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Lpg3ufQ3nhGNw1P2ia2R83"}]}]}} {"custom_id": "Sunamganj_35397_12312_35910_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3gnzx1dFXr2r5Qk5wjeBvK"}]}]}} {"custom_id": "Comilla_8208_9747_8721_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KKGZDRQuGuCFbd1m9zaXw1"}]}]}} {"custom_id": "Tangail_11799_31806_12312_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3s78njeBif2So7s8wRy3Am"}]}]}} {"custom_id": "Mymensingh_33858_33345_34371_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BSWZrYdztv9gZVkNEpGJqK"}]}]}} {"custom_id": "Noakhali_13338_45144_13851_45657.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3ozFhoGpp5heSxbCfPtM3o"}]}]}} {"custom_id": "Gazipur_6669_7695_7182_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WTvNajX2tCk6Do8QHvH72J"}]}]}} {"custom_id": "Pabna_29754_22572_30267_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gd6hU1Wgtb5Z1MX9jm8ivL"}]}]}} {"custom_id": "Panchagarh_1026_8208_1539_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1adsCt4Tzq7gyCifwxaYa4"}]}]}} {"custom_id": "Lalmonirhat_1026_1026_1539_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PvrF88DPMAD1Y3Zx69PsvE"}]}]}} {"custom_id": "Thakurgaon_11286_4104_11799_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KrBuMuN7SaJfd1Nadjk6qX"}]}]}} {"custom_id": "Kushtia_22572_18981_23085_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J35rhRC62TTAyVStLBHHcf"}]}]}} {"custom_id": "Munshiganj_24624_6669_25137_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Y7RhTi91BGz3AkxeYCNkTd"}]}]}} {"custom_id": "Narail_9747_10773_10260_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P2yikaVDUZtRn3de6H1LSE"}]}]}} {"custom_id": "Khulna_13851_38475_14364_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9msdHPCrpYLT3Q49fjrRGw"}]}]}} {"custom_id": "Natore_12825_1026_13338_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KdJrn9Jt1qgjspdMTrfAc1"}]}]}} {"custom_id": "Dinajpur_2565_11799_3078_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NsZxz5LDTu6s2YqnGrMzp4"}]}]}} {"custom_id": "Rangamati_28215_73872_28728_74385.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HtPWGw77payQJ93xkNPLHG"}]}]}} {"custom_id": "Bogra_26676_10260_27189_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8uMsyzMcbodSioAdWPqTJX"}]}]}} {"custom_id": "Habiganj_10260_9234_10773_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Pf6hDxorucXMXcD8RAA5aw"}]}]}} {"custom_id": "Rangamati_28215_70281_28728_70794.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LasDFhTMdLSqjPq1JXoJKh"}]}]}} {"custom_id": "Barguna_6156_17442_6669_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KE36YiYQshh25edi5JLwVC"}]}]}} {"custom_id": "Narayanganj_6156_10260_6669_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TnkTN9rcYXZtMZioofCfDf"}]}]}} {"custom_id": "Chittagong_17955_6156_18468_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4axBubhuswcfjp57WD2h5C"}]}]}} {"custom_id": "Chittagong_13338_9747_13851_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EZD1bEcca6o62jhDXupthd"}]}]}} {"custom_id": "Satkhira_14877_28215_15390_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-URtarz8mmAq83YsA2pmnfv"}]}]}} {"custom_id": "CoxsBazar_12312_21033_12825_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6y86qXjNMjoa121hAi8Qeq"}]}]}} {"custom_id": "Habiganj_9747_5130_10260_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Lvoz5cNuEi5TqD5nXSAiDL"}]}]}} {"custom_id": "Chittagong_20520_27189_21033_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2L1yYhHaVKZ7jveCAtEtQJ"}]}]}} {"custom_id": "Naogaon_33345_29754_33858_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R56BmpDnWsx8wmeGFQjCDg"}]}]}} {"custom_id": "Netrakona_11286_8208_11799_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8wc848XDu5DzNvG17Qas6V"}]}]}} {"custom_id": "Bogra_28215_5643_28728_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NC7SP6HLgSWJwGi5AjziWc"}]}]}} {"custom_id": "Sirajganj_9747_26163_10260_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UcmGMJ6Xct9v3mJE3SvRJv"}]}]}} {"custom_id": "Maulvibazar_10773_26676_11286_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WYxaDtw9ZuNGHmNQj69T6G"}]}]}} {"custom_id": "Bagerhat_4617_21033_5130_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JwBv7oo1AHr2FUznamxyiY"}]}]}} {"custom_id": "Kushtia_23598_12312_24111_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Wga8mfazeGpz5RHNpnfGW2"}]}]}} {"custom_id": "Madaripur_1026_16929_1539_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MtDC1yXdJaesryEuG1eYF2"}]}]}} {"custom_id": "Comilla_9234_14877_9747_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FEo3K8NroqnmpQkw7neiBK"}]}]}} {"custom_id": "Satkhira_7182_52839_7695_53352.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N6Ss2hqipWqh72LfJ6qGHJ"}]}]}} {"custom_id": "Thakurgaon_8208_21546_8721_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KcjVkmVPDfUrscRAxQb4wH"}]}]}} {"custom_id": "Lakshmipur_5643_13851_6156_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9HsMbjyC3gUKXPo236obTk"}]}]}} {"custom_id": "Gaibandha_15390_2052_15903_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8FtwEq4waSBxakjGR2ZkXh"}]}]}} {"custom_id": "Jamalpur_9234_20007_9747_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EcDqVjsiVv9xPkBTMZLDxG"}]}]}} {"custom_id": "Chittagong_31806_29241_32319_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L8fkVSokZfiJkjktio8q1M"}]}]}} {"custom_id": "Khulna_11799_32319_12312_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EFZHPGsLyeYbsjMTvnYDZo"}]}]}} {"custom_id": "Noakhali_17442_19494_17955_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PyMunGuvqxAERwK2PQzSrn"}]}]}} {"custom_id": "Habiganj_2052_21033_2565_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TTp8ec6CkeqG6Xk1hqmidV"}]}]}} {"custom_id": "CoxsBazar_10773_36936_11286_37449.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5f7FFW5hmz7e9fLD2jvX71"}]}]}} {"custom_id": "Sunamganj_12312_22059_12825_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VTF5NNUxE7o989RxV5ZeH1"}]}]}} {"custom_id": "Manikganj_6156_3591_6669_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DPFjxRdj16U1mfktPy4wuh"}]}]}} {"custom_id": "Jhalokati_5130_14877_5643_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2KVvs9ggW9tropeVyMStB8"}]}]}} {"custom_id": "Magura_8208_10773_8721_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J6frNTVsiS9iV7Am3rVUdE"}]}]}} {"custom_id": "Rangpur_15903_24624_16416_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LDqacuhHfuKLsoP1CBKZDu"}]}]}} {"custom_id": "Natore_5643_12312_6156_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-THCcHoF5xdn3s2Y15cA6nj"}]}]}} {"custom_id": "Bandarban_8208_42066_8721_42579.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-84GiF1cUoSvhpZwTUvGvPs"}]}]}} {"custom_id": "Sirajganj_14364_32832_14877_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5GxCdV9Ae1AqqfpX1okozJ"}]}]}} {"custom_id": "Tangail_15903_23598_16416_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Eon6zLX1nSHeA533CTdQ2r"}]}]}} {"custom_id": "Nilphamari_13338_15390_13851_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TZHDWK8bjqCqRLZqL53u1q"}]}]}} {"custom_id": "Rajshahi_6669_14364_7182_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PestjV2KedpSP5CzGExD8f"}]}]}} {"custom_id": "Rajbari_17955_12825_18468_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Csu7cgBz7aXXzuq4kaCeRM"}]}]}} {"custom_id": "Brahamanbaria_16929_12312_17442_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FDWUxJDo2hft9XyXTJ2fsN"}]}]}} {"custom_id": "Naogaon_21546_3078_22059_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CLTwMtMAf16ZLe8tWBXgT3"}]}]}} {"custom_id": "Rangamati_28215_70794_28728_71307.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5BfWCGRyusXFKrW85HzzSN"}]}]}} {"custom_id": "Sunamganj_25650_11799_26163_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GrpuM6BCer3mpcSb8UwhhT"}]}]}} {"custom_id": "Mymensingh_21546_3078_22059_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RDGHBiTdfC5tU8SnqaFp3C"}]}]}} {"custom_id": "Khulna_12312_7182_12825_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ChUrYd6aQZXAfbLDuVaAJ1"}]}]}} {"custom_id": "Thakurgaon_1026_12825_1539_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R8zVpskYgLoim8oCjFVuvu"}]}]}} {"custom_id": "Naogaon_24111_9747_24624_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WyYWf7TEqrgnGTAWmLAK3E"}]}]}} {"custom_id": "Comilla_15903_17442_16416_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TipeVHuzbnxuPNUuaShMKP"}]}]}} {"custom_id": "Bhola_8721_28215_9234_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T5nsVRfEw8bkpx15TrBZct"}]}]}} {"custom_id": "Rajshahi_20520_8208_21033_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YZbrsTb2Z8yjnRurPEg8hj"}]}]}} {"custom_id": "Brahamanbaria_14364_20007_14877_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WyznsecR3J57ba8g9pgqCb"}]}]}} {"custom_id": "Mymensingh_26163_47709_26676_48222.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-73xQojuemmwVNz3H2ji332"}]}]}} {"custom_id": "Natore_13338_7695_13851_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6Zx8WmoRsG4g2QeL9zocvA"}]}]}} {"custom_id": "Khulna_8721_10773_9234_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-St4zi1iJU9HqesYistyRyP"}]}]}} {"custom_id": "Satkhira_15903_16416_16416_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BbW5GqBUEmQ7nwge52qABT"}]}]}} {"custom_id": "Tangail_7182_33858_7695_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QTfzme1ggi7xBHkPrZ1cXk"}]}]}} {"custom_id": "Noakhali_14364_9747_14877_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2qN2gKxTBspfhV5X6KXVMc"}]}]}} {"custom_id": "Jhenaidah_1026_24111_1539_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-57oy5QXhB7mKsFqmxyXV3D"}]}]}} {"custom_id": "Barisal_21546_14877_22059_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RmqnRnH2U5GHeMApxvmzQy"}]}]}} {"custom_id": "Tangail_22572_25137_23085_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FWz9d3HRVbe9ALtttENQdq"}]}]}} {"custom_id": "Kushtia_9234_8721_9747_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4AW6iW1eitLnAbC5uSH9ii"}]}]}} {"custom_id": "Khulna_6156_53865_6669_54378.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9QettUtszvapzbzu7YCCzA"}]}]}} {"custom_id": "Jessore_3591_13851_4104_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X5nDo41k7U8uEyVHHKgmgt"}]}]}} {"custom_id": "Rangamati_18981_52839_19494_53352.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4fiVdKoPHGfNMzbkxjozPH"}]}]}} {"custom_id": "Patuakhali_19494_36936_20007_37449.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NkHJvygCUZ2owG4tjJkE7q"}]}]}} {"custom_id": "Noakhali_8208_39501_8721_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bo7TNSWrnuBSQyPcR1zWWJ"}]}]}} {"custom_id": "Chittagong_35910_42066_36423_42579.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CubQFcfPTWXwpnkNqx22Mp"}]}]}} {"custom_id": "Khulna_10773_46683_11286_47196.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2d6rZYL4UZuB2bid1HGmE2"}]}]}} {"custom_id": "Chittagong_35910_29241_36423_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UuCqTZiFQNxEgqhfvc95qS"}]}]}} {"custom_id": "Barisal_3078_3078_3591_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TmffpJR18ovSmrxEzLQvcC"}]}]}} {"custom_id": "Nilphamari_10773_18468_11286_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Lz3Kh4qygeyeyUvwMEYR2k"}]}]}} {"custom_id": "Rangpur_15903_10773_16416_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RHPBQJMbS9zYBmgWQUBBLc"}]}]}} {"custom_id": "Bagerhat_6669_34371_7182_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FckCZG9ZNhFF5Y2jrtT8NU"}]}]}} {"custom_id": "Habiganj_13338_29754_13851_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7xwQq6uCR7tMLMCF4FEJCJ"}]}]}} {"custom_id": "Noakhali_11286_33345_11799_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EcYeLcWLmc1SR8rsk7jmD3"}]}]}} {"custom_id": "Narayanganj_13338_6156_13851_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QHZCwuZ6GMs9uEKsgBfr1L"}]}]}} {"custom_id": "Panchagarh_4104_7182_4617_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-72KyREueNmMxSaSs5XXjBU"}]}]}} {"custom_id": "Noakhali_15390_16929_15903_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F8cCTbijrsihi9SX6BRiJa"}]}]}} {"custom_id": "Thakurgaon_11799_2565_12312_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CUXkEoWVyUgg9MTzvCEmZi"}]}]}} {"custom_id": "Jamalpur_8208_21546_8721_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7rvnWotnpkLr3AEmDPk2ps"}]}]}} {"custom_id": "Khagrachhari_12825_23598_13338_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UmhahVZxKZeoubyiytnbnh"}]}]}} {"custom_id": "Bandarban_17955_29754_18468_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WVMCuboTpNSo2f6oBjez5K"}]}]}} {"custom_id": "Khulna_1026_24624_1539_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GRNwYnC5qakUd8CQoqVEBD"}]}]}} {"custom_id": "Khagrachhari_5130_25137_5643_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GwcHG626vH51qjbBKcFc3P"}]}]}} {"custom_id": "Satkhira_14364_34371_14877_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EdZqQ8n4nKkfvTcx21jprE"}]}]}} {"custom_id": "Gaibandha_11286_27189_11799_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VPKVzuugz4djkUPUgEbQ2T"}]}]}} {"custom_id": "Rajshahi_12312_4617_12825_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NDDD3Mr72yXqMpyfnzQa7V"}]}]}} {"custom_id": "Panchagarh_2565_5130_3078_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Le42SN2nvzbCbMErCURwpN"}]}]}} {"custom_id": "Khagrachhari_13338_4104_13851_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BNehxp5FHsrU4Hpd9kfaSx"}]}]}} {"custom_id": "Kurigram_17442_11286_17955_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NsEGb8FaEdVfW9uobXYHMu"}]}]}} {"custom_id": "Netrakona_10260_13338_10773_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-E2WrEbn9iKxH31Pt4sSi7a"}]}]}} {"custom_id": "Sunamganj_15903_29754_16416_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1FGFnWLE7yZiwXNnzJcXZw"}]}]}} {"custom_id": "Habiganj_16416_12312_16929_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6P3yjzjzWhyafMZp6c76BP"}]}]}} {"custom_id": "Dinajpur_22572_16929_23085_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Sx5yjV9sFJjSEmoZgp7fGA"}]}]}} {"custom_id": "Gazipur_14364_3591_14877_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CfXX5v9oX9JJ6MMBWNfrpj"}]}]}} {"custom_id": "Habiganj_13851_15903_14364_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LaG12RVknQkKfY25LKaRsz"}]}]}} {"custom_id": "Rangpur_21546_8721_22059_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-488mpNU6hfPd7iXzCaGpeD"}]}]}} {"custom_id": "Naogaon_32832_34371_33345_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ELu8zNUykGYYzEr1KbPk9J"}]}]}} {"custom_id": "Meherpur_3078_15903_3591_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4voT8BxkHbc2GJwJAFYooG"}]}]}} {"custom_id": "Comilla_30780_23085_31293_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TLyRtBsaeT6myQM5qvnsbL"}]}]}} {"custom_id": "Magura_13338_18468_13851_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SYGcRQsDhaRSvDEHUZuBZt"}]}]}} {"custom_id": "Rangamati_10260_23598_10773_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1mVxNL9aFhk3W7oVombq9b"}]}]}} {"custom_id": "Khulna_19494_12825_20007_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Jyx5zS4LiQskZRtb7rTiTh"}]}]}} {"custom_id": "Gaibandha_20520_28215_21033_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HX1Bgjskww1GxLaDQumtWr"}]}]}} {"custom_id": "Kushtia_5130_13338_5643_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XY9XAvYqW2o5cvCXsLp7Eg"}]}]}} {"custom_id": "CoxsBazar_20007_51300_20520_51813.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6cqnex7EY8fekNTfS84Gun"}]}]}} {"custom_id": "Kishoreganj_22059_17955_22572_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LG3VovrQ4cjdySNE9DmJMH"}]}]}} {"custom_id": "Sylhet_23085_7182_23598_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ThbMjSKgJC9tLkZH6iYFUD"}]}]}} {"custom_id": "Bandarban_20007_16929_20520_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T5vNAMNyiLA2dTBgai1uqA"}]}]}} {"custom_id": "Narsingdi_7182_5130_7695_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Xq2FR4mYDAsc2zuP7mSUU1"}]}]}} {"custom_id": "Gazipur_9234_10260_9747_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Pdn4YSEE6pHQwGJRVHkjfm"}]}]}} {"custom_id": "Khulna_12825_41040_13338_41553.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mwx5YVp8njP2bKhRWZmXK6"}]}]}} {"custom_id": "Kushtia_11286_11286_11799_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FZ4N2tXQNHUBnG9jGd5kke"}]}]}} {"custom_id": "Khulna_14364_37962_14877_38475.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BtCpXiaQMS4asV5wtGKJZn"}]}]}} {"custom_id": "Khulna_6669_26676_7182_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Pn5Gnkb1n6sDgwodsuXoaz"}]}]}} {"custom_id": "CoxsBazar_10773_2565_11286_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mhv71WfQivAjjZ1RD2SH87"}]}]}} {"custom_id": "Madaripur_8208_3591_8721_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8hvzhhKuTJ8eL5UUR5hjtr"}]}]}} {"custom_id": "Chittagong_24624_29241_25137_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QypBArqjdaNYn8BV978GSq"}]}]}} {"custom_id": "Habiganj_22059_1539_22572_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S2FU6Nf81yReist77NzwZC"}]}]}} {"custom_id": "Pabna_13338_15390_13851_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5rmC3Z8AJ1oAF6aSHpdSa4"}]}]}} {"custom_id": "Sunamganj_1026_3591_1539_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DgsRHSqi931NyJkwYdiMsb"}]}]}} {"custom_id": "Dinajpur_18981_19494_19494_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6tqA8x5B6NoVfbkv1wUBGD"}]}]}} {"custom_id": "Sunamganj_24111_9747_24624_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5gr6AyKiFd5NGP2gt88ZUK"}]}]}} {"custom_id": "Madaripur_18468_20520_18981_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-118u8SkJP6QVoMdA8h22x1"}]}]}} {"custom_id": "Barisal_20520_3591_21033_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LVJciAHKRQZHqv8WSztzoh"}]}]}} {"custom_id": "Gaibandha_13851_4104_14364_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GoaNtt68i1NoUxuMrvRdvz"}]}]}} {"custom_id": "Naogaon_28728_8721_29241_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9EgVdZM8u669nXTEZxiRua"}]}]}} {"custom_id": "Patuakhali_15390_7182_15903_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AfRgepdvKJeKN7sEFo3k3E"}]}]}} {"custom_id": "Narail_5643_11286_6156_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hot6FXKoW7Sds8kkGRdyZQ"}]}]}} {"custom_id": "Jamalpur_11286_34371_11799_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8kWBi95VasPWX4kBLPPk5g"}]}]}} {"custom_id": "Gaibandha_8721_21546_9234_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SVVjqkjS65aXQwgPUNwRGY"}]}]}} {"custom_id": "Tangail_10773_31806_11286_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Qip5jtQL6NcUVWrwfRhfjt"}]}]}} {"custom_id": "Patuakhali_17955_30267_18468_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QcS6nNyfeeoTY163Qfx9ua"}]}]}} {"custom_id": "Gopalganj_20007_16416_20520_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NqcveYyB1V6u29suUE7tXh"}]}]}} {"custom_id": "Comilla_28215_25137_28728_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UvPYu18dLLHFJQZqE9k8f6"}]}]}} {"custom_id": "Khagrachhari_7695_33858_8208_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6bY1cPo75Hyi5Xc9HzRa26"}]}]}} {"custom_id": "Kishoreganj_9234_8721_9747_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-55HKw8YnGHEWGKF8GjE24W"}]}]}} {"custom_id": "Panchagarh_12825_10260_13338_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LNs7Fi87b6i2tgcgHQoYes"}]}]}} {"custom_id": "Comilla_6156_17442_6669_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F2G8eSYsZyBQSPd1KVxMyB"}]}]}} {"custom_id": "Habiganj_5643_28215_6156_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AT5CeCUnyGGGb9UzSyqurc"}]}]}} {"custom_id": "Habiganj_12825_11286_13338_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-15rsQs7QhhkkUSNjnpWTrR"}]}]}} {"custom_id": "Mymensingh_31806_28728_32319_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PXhUxCagj7SLocoAuhQvso"}]}]}} {"custom_id": "Pabna_4617_15903_5130_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9DV1gQMdyDPyup2aF6MPrx"}]}]}} {"custom_id": "Sirajganj_11799_18468_12312_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GQWQvMjoom6VMejC2ehCAa"}]}]}} {"custom_id": "Kurigram_19494_14877_20007_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6eRamU2jQQVzDKug1AHE8f"}]}]}} {"custom_id": "Meherpur_0_13851_513_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T6Je4ERcyDC119PErv9i5B"}]}]}} {"custom_id": "Chittagong_27189_37449_27702_37962.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JaJTGwmSLQjr1nMMqT4J4S"}]}]}} {"custom_id": "Narsingdi_12312_7182_12825_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S7ftLmd8v8oJNenNRwG9KL"}]}]}} {"custom_id": "Comilla_12825_7182_13338_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2Kqi2PE8crAqHhjQfxtVam"}]}]}} {"custom_id": "Munshiganj_11286_2052_11799_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3pysw3RUJQ7iF2x28XcuBw"}]}]}} {"custom_id": "Netrakona_14364_25650_14877_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vth17ikziw5HGwdsD2hgz3"}]}]}} {"custom_id": "Bagerhat_4104_11286_4617_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SzyRtMEm6soy6V6BVW4Hsu"}]}]}} {"custom_id": "Brahamanbaria_18468_29754_18981_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1k3Toc6n6EVrnYrHR3NxJS"}]}]}} {"custom_id": "Khagrachhari_7695_51813_8208_52326.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R9M8Snv2HFwvzQWiiBEHys"}]}]}} {"custom_id": "Mymensingh_23085_31806_23598_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7WkHR1Yobvo73Lbb4RatfU"}]}]}} {"custom_id": "Shariatpur_3591_1026_4104_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XWYST68jKwKWhqUMUxkGip"}]}]}} {"custom_id": "Khulna_5643_33345_6156_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L3AsVJuMGXomWLZL8G2QiM"}]}]}} {"custom_id": "Pirojpur_1539_24624_2052_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GgfLDmtif2r4PFK27Mck8E"}]}]}} {"custom_id": "Bandarban_18981_33345_19494_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8xCbsvRPkQxpU5UxdBRtCR"}]}]}} {"custom_id": "Sirajganj_13338_27702_13851_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LBVi8Qmq493KGZXKGtoCFc"}]}]}} {"custom_id": "Naogaon_2565_6669_3078_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XxYVK9DcmVe1cf7V7aNL9v"}]}]}} {"custom_id": "Kurigram_15903_32832_16416_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VgR5NLDrSkhr32bv7SU3b5"}]}]}} {"custom_id": "Satkhira_9234_3591_9747_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AS7p9i6ATm6AwSdwKPXzhc"}]}]}} {"custom_id": "Barguna_4617_21033_5130_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DVJVnucDX3MJZMbu72NjwA"}]}]}} {"custom_id": "Khagrachhari_13851_17442_14364_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MP24brzGs46xGkgqZvxNWR"}]}]}} {"custom_id": "Kurigram_8208_29241_8721_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CmukgyYX917pNX79en449v"}]}]}} {"custom_id": "Chuadanga_7695_21033_8208_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-57izR8qmecjmMYouEW2ErY"}]}]}} {"custom_id": "Faridpur_26163_19494_26676_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CKxJLjKVK34sAC9hBZyGU5"}]}]}} {"custom_id": "Sirajganj_21546_30267_22059_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MtBK22gqMyvbSJZxRT5avZ"}]}]}} {"custom_id": "CoxsBazar_16416_40014_16929_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QxYpgEb4gE5bVkbz1ZxWWB"}]}]}} {"custom_id": "Noakhali_9234_31293_9747_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6igTrjB24DLv3fByNmvaGJ"}]}]}} {"custom_id": "Gopalganj_19494_12312_20007_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NqsEx1oS1CzcNtiqpHEMDu"}]}]}} {"custom_id": "Bandarban_11799_3591_12312_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UsHsACXQZD7DDugXepWogu"}]}]}} {"custom_id": "Dinajpur_28215_34884_28728_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1g2HDXf1urWEBD22Ksnr44"}]}]}} {"custom_id": "Noakhali_6669_41040_7182_41553.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TRyYY3SkwitmUC2uTsknS2"}]}]}} {"custom_id": "Bandarban_9747_26163_10260_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2wEhW4sNj7pCMZCx43tAky"}]}]}} {"custom_id": "Noakhali_11799_6156_12312_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4tQ2zRYurdsArgGBDpneho"}]}]}} {"custom_id": "Maulvibazar_28728_4104_29241_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QDZ48UVRkNR3DW441J5Gje"}]}]}} {"custom_id": "Bagerhat_6669_33858_7182_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TuHzKQ58RdbRNvzPwbFAKr"}]}]}} {"custom_id": "Kurigram_16929_21546_17442_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2Rh53fQFpAYevMsQV7sYKM"}]}]}} {"custom_id": "Narayanganj_4617_3591_5130_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Emjfbyu6s4a3XpuAwBXPaN"}]}]}} {"custom_id": "Munshiganj_17955_8721_18468_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YLVgR332Eoj5bmSTgcZoR2"}]}]}} {"custom_id": "Brahamanbaria_16416_18468_16929_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HnFPGur1qAaRE34g1ekWL2"}]}]}} {"custom_id": "CoxsBazar_1026_7695_1539_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Lo398bM9RsiwvwWte39Cue"}]}]}} {"custom_id": "Jessore_22059_12825_22572_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4tSzWW4FTsCfFzhHR8kaCx"}]}]}} {"custom_id": "Sherpur_513_14364_1026_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QHuPnbC5rmhEfxijZ4NAfq"}]}]}} {"custom_id": "Netrakona_8208_14877_8721_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R4r776Yzj2Vb56EyJqitEm"}]}]}} {"custom_id": "Gaibandha_7695_17955_8208_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-713rAhqwaL4AsYekCfTiDg"}]}]}} {"custom_id": "CoxsBazar_10773_27702_11286_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3d4CBujvsYvJwdKfue97WU"}]}]}} {"custom_id": "Panchagarh_4104_5130_4617_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X5D6LDtk2QJcMqyxncqEDn"}]}]}} {"custom_id": "Habiganj_6669_16929_7182_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9NJpqLc6vxsCKtL9tYRTu6"}]}]}} {"custom_id": "Khulna_5130_33345_5643_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RQZTYLYGxxrYuJGgxKk72d"}]}]}} {"custom_id": "Magura_13851_17955_14364_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HeBVR8CA6QL9qdB9uqCrJw"}]}]}} {"custom_id": "Jessore_10773_20007_11286_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bu7HC1Cg5tz4jPSetYHSLK"}]}]}} {"custom_id": "Jessore_2565_17442_3078_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SbkRLz71fUkunDDxvraJsa"}]}]}} {"custom_id": "Bandarban_6669_7695_7182_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HzUSCYZ74z8YEzm6uy7ZAn"}]}]}} {"custom_id": "Sirajganj_21033_5643_21546_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RJJGw2RSDBmzWPHZMcEEDg"}]}]}} {"custom_id": "Shariatpur_13851_15390_14364_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Q6geEmf2CniifX8rtDnDgf"}]}]}} {"custom_id": "Satkhira_5643_4617_6156_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8dsQecdFuRB357kTCxEP7a"}]}]}} {"custom_id": "Comilla_17955_13851_18468_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PQ4zoqMKfmpzvNeAeMZqWg"}]}]}} {"custom_id": "Gazipur_23598_11799_24111_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HJCR6DJTUivKWEhMR3mDK8"}]}]}} {"custom_id": "Jamalpur_2565_14364_3078_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QLJCxcnFAd7Hx6FQMsvaeP"}]}]}} {"custom_id": "Gaibandha_18468_21546_18981_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S5xe2NUfrMQtc5GpokDSfA"}]}]}} {"custom_id": "Dinajpur_11286_3591_11799_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1JDfZESygdKbB6CB161Wp8"}]}]}} {"custom_id": "Jessore_16929_24111_17442_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1rgTiG7WKwjHL4TBVExojT"}]}]}} {"custom_id": "Netrakona_21033_7695_21546_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GwMtqADLdgaAcHJs78KrKP"}]}]}} {"custom_id": "Barisal_2565_11286_3078_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VS5iLoUdGkGiAjaPYkPxxQ"}]}]}} {"custom_id": "Kushtia_24111_22059_24624_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JhoEjvHe3RqhpASFi6EPrp"}]}]}} {"custom_id": "Pabna_10773_10260_11286_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JkfBGhVTD42v1u1NgfJG4H"}]}]}} {"custom_id": "Bhola_17442_37962_17955_38475.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U2Ff5ufnGcLd6KKhJVeDoL"}]}]}} {"custom_id": "Khagrachhari_8721_20520_9234_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1A3NBRJTqmAKQErZQXj9nj"}]}]}} {"custom_id": "Madaripur_14364_13338_14877_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GajgXSPC5wKvt9E4bSCrQH"}]}]}} {"custom_id": "Barisal_12825_12825_13338_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JJYzMg5dbHeFYstFatRJK5"}]}]}} {"custom_id": "Shariatpur_7182_4617_7695_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DxsXr1CGkX6UX7gotXAfDM"}]}]}} {"custom_id": "Rajshahi_4104_16416_4617_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ce8BymeUqjTECBF3wCFbE6"}]}]}} {"custom_id": "Habiganj_22572_27189_23085_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6H6yP9kApFyEv71cgLyG6P"}]}]}} {"custom_id": "Rangpur_11286_16929_11799_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6WDMCdoM8m5CQAj3YFsXrd"}]}]}} {"custom_id": "Chandpur_12825_10773_13338_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4hjF3hVDEfgFPkNwRfpQtQ"}]}]}} {"custom_id": "Sylhet_31806_10773_32319_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W88nTbrWb5xXyxRGYFyRZF"}]}]}} {"custom_id": "Rangpur_16416_21546_16929_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YLLUbzJdWx1xgA9eeCPuwV"}]}]}} {"custom_id": "Magura_3078_13851_3591_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-79zTMUpaWaArd9X92gqTJu"}]}]}} {"custom_id": "Sherpur_2052_13851_2565_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WaXce6XK28jmEX4poXpdSH"}]}]}} {"custom_id": "Patuakhali_12825_17955_13338_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YNGNKfak7ecnak3ufdSteT"}]}]}} {"custom_id": "Natore_4617_24111_5130_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-B5TYExiERkU8aeW1Zhn6He"}]}]}} {"custom_id": "Dinajpur_6156_11799_6669_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W7PxRWqEtndtKHCwcyAsrV"}]}]}} {"custom_id": "Lalmonirhat_3591_1026_4104_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QEdDUNWteLFqX9TxsCGghk"}]}]}} {"custom_id": "Chandpur_14877_6669_15390_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RBQTotF7q4tafjJcvPrsyC"}]}]}} {"custom_id": "Jessore_16416_11799_16929_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AyerYYpgi2FMfGEkCGE3Va"}]}]}} {"custom_id": "Tangail_8721_8721_9234_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AbdgNBZpTD46aPJchcaqBu"}]}]}} {"custom_id": "Pabna_25137_22572_25650_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KkYWi4eZvzFbW7H6geHJMP"}]}]}} {"custom_id": "Barisal_16416_8721_16929_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M9KQRTdhw5HyBSARrjnsJK"}]}]}} {"custom_id": "Lakshmipur_13851_16929_14364_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XuVCyEwoawyq269WEQaFuC"}]}]}} {"custom_id": "Bagerhat_15390_26676_15903_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FBngdeHFHBaxXpZCHgkm9h"}]}]}} {"custom_id": "Faridpur_26676_16929_27189_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BkmwtPxmcDXSmsdLjxxrSt"}]}]}} {"custom_id": "Rangpur_5130_11799_5643_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5Ne6eX22nXsHJGyRmPMsVf"}]}]}} {"custom_id": "Kurigram_14364_27189_14877_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-C27q7GS91NCTCKD16NorbQ"}]}]}} {"custom_id": "Manikganj_17955_8208_18468_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ncr2qvJtVvcw7j882kqfdx"}]}]}} {"custom_id": "Pabna_19494_15903_20007_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UkcSZLkaDRbzUh7cidXPwo"}]}]}} {"custom_id": "Faridpur_25137_7695_25650_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JPcTYPSDyR8kCFGZN61ifo"}]}]}} {"custom_id": "Comilla_1026_11286_1539_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TCGn2RqwCL8PLjVFVRmXqH"}]}]}} {"custom_id": "Sunamganj_16416_21546_16929_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MHSx7LrtAh57aAxqJL4bhu"}]}]}} {"custom_id": "Rajshahi_18981_7182_19494_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RHd36MBsaS9aDhEmbjathz"}]}]}} {"custom_id": "Pirojpur_0_8208_513_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MhNHQc3CiR7FYddrMhfeC1"}]}]}} {"custom_id": "Khulna_3078_25650_3591_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RRQQznZKFGnSWXuBoomgo9"}]}]}} {"custom_id": "Habiganj_21546_24624_22059_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1CQ96LCUoUgG27znvKiGMx"}]}]}} {"custom_id": "Sunamganj_13338_15390_13851_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TawaiWZd3cu92mWm57m8wE"}]}]}} {"custom_id": "Pabna_10773_9747_11286_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YXa2dmQ5Bn1gzM6u4TkTR5"}]}]}} {"custom_id": "Jamalpur_9747_13338_10260_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4ETvGjh1z97y2nGG5bDkXJ"}]}]}} {"custom_id": "Gaibandha_13851_17955_14364_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NLrPnReHmiBEoLKnTTFHg3"}]}]}} {"custom_id": "Pabna_22572_9234_23085_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XZqWhCEC5oHzpsVCfVLBzn"}]}]}} {"custom_id": "Satkhira_19494_11286_20007_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1uGqTPXqqh4gJj1NsWgNwj"}]}]}} {"custom_id": "Jhenaidah_13338_8721_13851_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-89GfnroT8mmHS7T6yXeijh"}]}]}} {"custom_id": "Rajshahi_12312_4104_12825_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UyW5gmmvraVZSYtyRjxXpq"}]}]}} {"custom_id": "Magura_12825_20007_13338_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M31jG6Ln6mG1Kou8wqxEaJ"}]}]}} {"custom_id": "Brahamanbaria_22059_23085_22572_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U4shAZB5M6NqiZZu3uydLo"}]}]}} {"custom_id": "Nawabganj_18981_12312_19494_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YJauszJJvyLFktuTb8bjxr"}]}]}} {"custom_id": "Barisal_9234_5643_9747_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GX4pqGq6J22Y7XPPtiTyhF"}]}]}} {"custom_id": "Narayanganj_14364_5643_14552_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2jRZ5ruVL3LSvEXKScDXu9"}]}]}} {"custom_id": "Naogaon_12825_9747_13338_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Rg46441LQx4ErHdS8mG7dW"}]}]}} {"custom_id": "Comilla_8721_10773_9234_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Tw61PGKbTPbYpxk4jnMvzW"}]}]}} {"custom_id": "Munshiganj_18468_8721_18981_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QYQKqrVzhA66ZnRAoUFW17"}]}]}} {"custom_id": "Faridpur_16416_11799_16929_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4h2dw7akn9xyGAbqDWiE8G"}]}]}} {"custom_id": "Sirajganj_21546_27189_22059_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7xsjYkQrrqpEyx84wZxcNM"}]}]}} {"custom_id": "Faridpur_14877_9747_15390_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DBX4NHyeMGezzdkrRUntHr"}]}]}} {"custom_id": "Rajshahi_12312_18981_12825_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MFcFY94HcRYVn222mKu9jd"}]}]}} {"custom_id": "Feni_8208_11799_8721_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Qt3d1XoxH6eB9AQZ9E7M7P"}]}]}} {"custom_id": "Barguna_9234_10773_9747_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Uyt4dEapzFUikrMddJpvzc"}]}]}} {"custom_id": "Dinajpur_32832_28728_33345_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NR11HRBtMJ9MEiUxiGWzK7"}]}]}} {"custom_id": "Comilla_15390_23085_15903_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8FNnsRLQ1YzM9NqPmfU99D"}]}]}} {"custom_id": "Khagrachhari_5643_30267_6156_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FACc9q6LjP1wqFRQfkqmbC"}]}]}} {"custom_id": "Nawabganj_11799_10773_12312_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2186PiV86cdYyQStrBiK2Z"}]}]}} {"custom_id": "Lakshmipur_2052_7695_2565_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QfUWSiAyp66N5PvepRA4hn"}]}]}} {"custom_id": "Bagerhat_10773_37962_11286_38475.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BAq3E412J2Ap6JqX2Up1Xn"}]}]}} {"custom_id": "Munshiganj_7695_8208_8208_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QC4i5PUFpB1FeWNhRkPhbm"}]}]}} {"custom_id": "Kurigram_13338_5643_13851_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LHdafyQD8hV1wG3rJRLsVf"}]}]}} {"custom_id": "Rangamati_9747_44118_10260_44631.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SN3q6ABMHjBE1FGhmdDB1C"}]}]}} {"custom_id": "Chandpur_1539_22572_2052_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6CtXyJujhgSgphLyEdXtrd"}]}]}} {"custom_id": "Bagerhat_10260_16929_10773_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NiQapr7i3dJhVEaXCs8uuJ"}]}]}} {"custom_id": "Thakurgaon_23598_9234_24111_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GhZAW15q2P2xTN2j2AxfBa"}]}]}} {"custom_id": "Patuakhali_22572_7695_23085_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2x9jtNjPFDZbwHPQLJ5whU"}]}]}} {"custom_id": "Rangamati_11286_1539_11799_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TiXrQNrgXkU1dbSwqHpAja"}]}]}} {"custom_id": "Dinajpur_14877_27189_15390_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2pNq5NxijZRKby34wj9euA"}]}]}} {"custom_id": "Panchagarh_20007_20007_20520_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BsJiu1MxCnj9MgLYgwgv9a"}]}]}} {"custom_id": "Mymensingh_9747_37449_10260_37962.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QerUKRh2rp7kmPQigbJ3Tc"}]}]}} {"custom_id": "Sherpur_1539_14364_2052_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2LrxygduGtAWBWWCz4wFow"}]}]}} {"custom_id": "Magura_12825_17442_13338_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3jTs5jPPAXxgW8uQZ2CKZZ"}]}]}} {"custom_id": "Bandarban_8721_33345_9234_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1pTU2iSw4QfQqzVp8awmUj"}]}]}} {"custom_id": "Rangpur_2565_12312_3078_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ts9uHnUShwJEnVNgbSGDTr"}]}]}} {"custom_id": "Bogra_24624_11286_25137_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hh6EqS1seGyjmZ666DsJqj"}]}]}} {"custom_id": "Pabna_20007_8208_20520_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N7vxCZCb3WKGvwJhuFs9NT"}]}]}} {"custom_id": "Sunamganj_9234_4104_9747_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-B4Y45zMvHdDqFcSAeVUohk"}]}]}} {"custom_id": "Gazipur_17955_7182_18468_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PSRe9M3KZZwJCnRHj1iFQS"}]}]}} {"custom_id": "Bagerhat_5643_33345_6156_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1KMBQ75AC9GkTDUc2BqbeM"}]}]}} {"custom_id": "Habiganj_10773_4104_11286_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8mssvviGMYZn5NeLKBwLyC"}]}]}} {"custom_id": "Pabna_19494_2052_20007_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MPErqXY6BWYXSjkcZeTDpy"}]}]}} {"custom_id": "Bhola_17955_15390_18468_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RZtvXXwczwyHKCd5fKCZji"}]}]}} {"custom_id": "Satkhira_11799_47709_12312_48222.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ArmKyUgLJ2YDSxJ7Dp6PwF"}]}]}} {"custom_id": "Feni_3078_11286_3591_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SzXA5VfLgW1FKNMHT4sfAs"}]}]}} {"custom_id": "Gaibandha_14364_28215_14877_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-15rXYrPJW6ifLxBas53XU8"}]}]}} {"custom_id": "Chuadanga_10260_2052_10773_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DDm7oHdHGukjFYxhm67RGU"}]}]}} {"custom_id": "Gopalganj_1539_13338_2052_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VyYRE2fzieGYe3pCT7aRZf"}]}]}} {"custom_id": "Mymensingh_0_20520_513_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-676pBVM8UQPkRcRcZKfAP2"}]}]}} {"custom_id": "Thakurgaon_15903_6156_16416_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CHSuf6UyFr3kbjVHQuuAF4"}]}]}} {"custom_id": "Kurigram_9747_3078_10260_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VAtkTeowppcLtcct9FZGyU"}]}]}} {"custom_id": "Satkhira_11286_27702_11799_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GZPLyWYYZ8SaduWM7B1bah"}]}]}} {"custom_id": "Sirajganj_11799_14877_12312_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FKoywXiz8SkE8MnaoqaQgK"}]}]}} {"custom_id": "Khagrachhari_15390_4104_15903_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fsf4VGZEWeXomNnqWNAnKy"}]}]}} {"custom_id": "Pabna_5130_12312_5643_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8wPB31ns5YpJ1m7K8aoaWd"}]}]}} {"custom_id": "Comilla_24111_18981_24624_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XxJfMUmMTNpUdeR8V53Ryq"}]}]}} {"custom_id": "Narayanganj_8721_7695_9234_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U6m4WFMjScV3RV34yJ6ypz"}]}]}} {"custom_id": "Jhalokati_1026_20007_1539_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AsWDLNUZLRRaduMYJhxLkX"}]}]}} {"custom_id": "Noakhali_9747_46683_10260_47196.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-McFA5k1H1F4s273ut6Xiyo"}]}]}} {"custom_id": "Habiganj_10260_26163_10773_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NK9ycTDKqAKsKKHBB1SM1y"}]}]}} {"custom_id": "Khulna_9234_29241_9747_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MWrSXsULbN6oT15L8CpHim"}]}]}} {"custom_id": "Rajbari_4104_8721_4617_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GW6AdgX87JoCubzN5AH2nV"}]}]}} {"custom_id": "Sunamganj_16416_24624_16929_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WKqyDkSGQELdRicW621GCL"}]}]}} {"custom_id": "Chittagong_22572_36423_23085_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QAwcD2dG4xpfHjScK7VyGx"}]}]}} {"custom_id": "Tangail_16416_35397_16929_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-E9i7VXokfBRVyJCzVyJ1yV"}]}]}} {"custom_id": "Comilla_23085_15390_23598_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Q6qM3QrTaxqQoMmS5J1jo8"}]}]}} {"custom_id": "Netrakona_30780_22059_31293_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PRYdv6jgnfLJDkYEctsihV"}]}]}} {"custom_id": "Pirojpur_3591_24111_4104_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KMob9K3h3X5UMQLJDQ1htT"}]}]}} {"custom_id": "Habiganj_23598_4104_24111_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JyhbxP12HMiuDnoEV5iL2B"}]}]}} {"custom_id": "Thakurgaon_15903_15903_16416_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RL2MGe5bM1dpdfj3bzTtax"}]}]}} {"custom_id": "Lalmonirhat_5643_12312_6156_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CaFLRg3LtbF1VKkibqEN1N"}]}]}} {"custom_id": "Lakshmipur_11286_2565_11799_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FxVqUPC78hgEnKyEekMh2D"}]}]}} {"custom_id": "Patuakhali_21546_11799_22059_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4EkCXaiiG5eqHVMazs4LTM"}]}]}} {"custom_id": "Jessore_16929_25137_17442_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-18JmaUgNqR5u1CpBmtrr6d"}]}]}} {"custom_id": "Khagrachhari_12312_22059_12825_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U74JtkoVuGMM8cDipkmjUz"}]}]}} {"custom_id": "Comilla_14364_6156_14877_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QjJzGmDcwTCvQ8pbEB6ENx"}]}]}} {"custom_id": "Nawabganj_6156_8208_6669_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WYeVQCerDyGjjxB1wzQDUp"}]}]}} {"custom_id": "Pabna_22059_14364_22572_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RgEqvTJVruhQ3fr9HzYrFp"}]}]}} {"custom_id": "Madaripur_16929_22572_17442_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CksNVieBy5hinTPRKw82Yg"}]}]}} {"custom_id": "Bhola_5130_18468_5643_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DhEiyKZLnoggBdbv3ycLUV"}]}]}} {"custom_id": "Khagrachhari_18981_24111_19494_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4rrGDgugwZUxwXBEjcG1XT"}]}]}} {"custom_id": "Shariatpur_8208_4617_8721_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XrEyxDUUDfbaU5GBMAVkGj"}]}]}} {"custom_id": "Rangamati_30267_85158_30780_85671.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Sn1nbysFgwuVsMedooCoaf"}]}]}} {"custom_id": "Khagrachhari_12312_32832_12825_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9aGXj3XKRnn4xTEdjBu5Gw"}]}]}} {"custom_id": "Satkhira_8721_54891_9234_55404.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LKVe5BTR9uSo72EnZCxzwS"}]}]}} {"custom_id": "Bagerhat_14877_17955_15390_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DyRz6UC3PC1rkkt37XFkUY"}]}]}} {"custom_id": "Bandarban_9234_3591_9747_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vp3riScXPQkxk6Zzbn64pB"}]}]}} {"custom_id": "Gazipur_11799_19494_12312_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5YrmfpXmxVZq1prHWWe2LR"}]}]}} {"custom_id": "Gopalganj_10260_13851_10773_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CG7qMvuZPizj3uPtp7q5GF"}]}]}} {"custom_id": "Lakshmipur_1026_7695_1539_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WvuUbBM4jiwvt1WMg9UhLX"}]}]}} {"custom_id": "Netrakona_10773_8721_11286_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FHiShw5wDsGcCGcxHogLaa"}]}]}} {"custom_id": "Bogra_27702_16929_28215_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Szf99Kb5tdUouvkFemi7oz"}]}]}} {"custom_id": "Netrakona_9234_6156_9747_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SateyZoH11fLcAZZpsWijb"}]}]}} {"custom_id": "Nawabganj_9747_13851_10260_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WnnDBxVENnbHjegqEa8VSK"}]}]}} {"custom_id": "Lakshmipur_14364_20007_14877_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Kuzqsnp92dB2JdkxuTxhV4"}]}]}} {"custom_id": "Mymensingh_29241_34884_29754_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6zW3GBacwmLyVfPmmCKmBC"}]}]}} {"custom_id": "Manikganj_513_3591_1026_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hrn4mKuNE21eMbphmbZDah"}]}]}} {"custom_id": "Narsingdi_3591_15390_4104_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4VuXuPYU7o2ZVV8GqJdvJh"}]}]}} {"custom_id": "Chittagong_26163_22059_26676_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V4JG3sJ4st9xcKJnFByFvU"}]}]}} {"custom_id": "Thakurgaon_6669_5130_7182_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CGNcqERXpE6waHvaMxMQHu"}]}]}} {"custom_id": "Sirajganj_24111_5643_24624_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2EmWqTAKmLs6P9KWZ41LQq"}]}]}} {"custom_id": "Natore_3591_19494_4104_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PwnjHri9S789uvYVCpKXuB"}]}]}} {"custom_id": "Rangpur_8721_10260_9234_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1gSEKBjTxyUeJgdMhJcPhA"}]}]}} {"custom_id": "Jamalpur_8721_40014_9234_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EbUdZNmZnVT4Kfg7bcgFyn"}]}]}} {"custom_id": "Gazipur_16416_21546_16929_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UZM7JWKVcFL99nw2b6VVd4"}]}]}} {"custom_id": "Madaripur_13338_7182_13851_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-B2TnkwB5Qwob38qDxeiUQi"}]}]}} {"custom_id": "Sherpur_13851_17955_14364_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KWjQ1tSdN7APgtaWL8nLrd"}]}]}} {"custom_id": "Rajshahi_6156_16416_6669_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V1FvvZxDwHNTncVfnMv1uc"}]}]}} {"custom_id": "Madaripur_6156_10260_6669_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Y3wPVmG7cAbUDwbhKCDZKA"}]}]}} {"custom_id": "Habiganj_5643_21033_6156_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WuYHDoEzsnYGqfyMPXg9Jx"}]}]}} {"custom_id": "Rangamati_14877_53865_15390_54378.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2kiDDHbGQBkCP3LVfZSppS"}]}]}} {"custom_id": "Maulvibazar_6669_23085_7182_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9QtbVfezUMe52nuq6Xk8np"}]}]}} {"custom_id": "Satkhira_17442_58995_17955_59508.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RrX36kS4rkbBqt1whfHxon"}]}]}} {"custom_id": "CoxsBazar_17442_25650_17955_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3rDPd5CEGdeFBRbwKQECVa"}]}]}} {"custom_id": "Naogaon_33858_25650_34371_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EKnRthweTV1KmJyMxA1vCb"}]}]}} {"custom_id": "Maulvibazar_21033_18981_21546_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RGASPBzAypQR8mvEJsWUsC"}]}]}} {"custom_id": "Bandarban_21033_30780_21546_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RXwEtc88ZV2prdNYtJNEoX"}]}]}} {"custom_id": "Jessore_18468_21546_18981_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4oTepbk98gEjnMi5nF4vhA"}]}]}} {"custom_id": "Habiganj_20520_1539_21033_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HQEYtmXkixjbyFhU5etPUU"}]}]}} {"custom_id": "Khagrachhari_7182_31293_7695_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5TpGc8vytLxSGJAGFN749d"}]}]}} {"custom_id": "Satkhira_16416_40527_16929_41040.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BD3RTrAdne1CHHCSHg3JwJ"}]}]}} {"custom_id": "Lalmonirhat_4104_3591_4617_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EfQzDZVFzJKMdaErMWessz"}]}]}} {"custom_id": "Sunamganj_17442_4617_17955_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1KKFpoUdZRatAMjY9DTHfH"}]}]}} {"custom_id": "Bandarban_20520_25137_21033_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ECcdxVrfTADZuPgjbhFZ2B"}]}]}} {"custom_id": "Satkhira_16929_40527_17442_41040.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X89UT3WBjmgqjfBzYDKmQz"}]}]}} {"custom_id": "Rajshahi_26163_10260_26676_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D7srh7M48xtjr4f6iUuNNY"}]}]}} {"custom_id": "Jhenaidah_25650_10773_26163_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GDBHTw4VP5TtaQyVnoeq9y"}]}]}} {"custom_id": "Natore_11286_8208_11799_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4EUNFM1ECeLQYQb8K4JCda"}]}]}} {"custom_id": "Rangpur_13338_20007_13851_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-G2XKMzuyLqbx8qpTfrkR6A"}]}]}} {"custom_id": "Satkhira_16929_35397_17442_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DaYRfum9rNK3ebNLmVNMJY"}]}]}} {"custom_id": "Bandarban_7695_15390_8208_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RzME9vVNcjxQ7LAGNVv77s"}]}]}} {"custom_id": "Rangamati_11286_30267_11799_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HxMjNDqgbkNxJoWgeJBwnn"}]}]}} {"custom_id": "Chittagong_26163_49248_26676_49761.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CER5TvJXwyzzg8y8D8Y2DE"}]}]}} {"custom_id": "Barisal_5130_1026_5643_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2ZHJbyd5bEUEysjcumEJ9b"}]}]}} {"custom_id": "Habiganj_14877_29754_15390_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DQ23ftBBbcc8BZYrTmwgHu"}]}]}} {"custom_id": "Gopalganj_9747_9747_10260_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DKTnu2nkavWGACbB1bs98k"}]}]}} {"custom_id": "Kushtia_4104_7695_4617_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KPwaaFFcERQ8C1A7mbXc2D"}]}]}} {"custom_id": "Jessore_16416_5643_16929_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JhsY2KqQhBFhK2f2cx3BA8"}]}]}} {"custom_id": "Naogaon_17955_26163_18468_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WDHJoSSdfcSC2TJaQoKCcX"}]}]}} {"custom_id": "Barisal_21546_26676_22059_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WThFKweej3tL54nJEs2Dc9"}]}]}} {"custom_id": "Natore_11799_15390_12312_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9VX5HNmc3qHVhi6brTJMG7"}]}]}} {"custom_id": "Sirajganj_22059_15390_22572_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KBAQURxMR2hUQYbnwdkgiu"}]}]}} {"custom_id": "Patuakhali_13338_24624_13851_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MmSHD7CyvgpxxyQscns9Mz"}]}]}} {"custom_id": "Rangamati_21546_24111_22059_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vj4MEjDM85ERwBd8JtTnfF"}]}]}} {"custom_id": "Chandpur_10260_17442_10773_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3gorYtMFvZejG4yF9nCxLB"}]}]}} {"custom_id": "Netrakona_7695_10773_8208_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JhSCZ6a39LhFNzE8rpFryH"}]}]}} {"custom_id": "Rangpur_9234_3591_9747_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ExL4hXdmYvV2FMyLWayo1e"}]}]}} {"custom_id": "Patuakhali_19494_4617_20007_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L1u1tEvqNuPqkwKY1a61g2"}]}]}} {"custom_id": "Rangamati_5643_50787_6156_51300.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-98EPzEmoJwg4E4qTTMCfJQ"}]}]}} {"custom_id": "Bhola_3078_14877_3591_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XyMpd5CgMvAwA3e2YcykAU"}]}]}} {"custom_id": "Jhenaidah_6669_26676_7182_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2TwH666fELfVmFeHrwXSgR"}]}]}} {"custom_id": "Narail_6669_14364_7182_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FJGHkpCkQa8Tk8afNoPfbx"}]}]}} {"custom_id": "Narsingdi_15903_15903_16416_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Y39MNJFPkXY5THzEz5KtBC"}]}]}} {"custom_id": "Nawabganj_9747_12312_10260_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NoXMyNCvQETFPcdSPUxaEK"}]}]}} {"custom_id": "Feni_6156_6669_6669_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5TtgoMk5fSYUx86RbqaZW8"}]}]}} {"custom_id": "Bandarban_7695_9747_8208_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FLW1mm79dzcs3FE4DYPqPq"}]}]}} {"custom_id": "Dinajpur_26676_29754_27189_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P6NFSbftMNzU5Wh4sNcFHq"}]}]}} {"custom_id": "Chittagong_36936_42579_37449_43092.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ka52UGx5AZih45AnznvKyH"}]}]}} {"custom_id": "Shariatpur_4617_15903_5130_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LAaKf7ch25GxZsr3KNou3d"}]}]}} {"custom_id": "Faridpur_28215_11799_28728_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LPMTdJ7sqB26cekW4Dh7Py"}]}]}} {"custom_id": "Bagerhat_7695_34371_8208_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3URvaqwKTS9kEAamCHKokf"}]}]}} {"custom_id": "Comilla_14364_13338_14877_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V51TLzQekM8d6L5s5to5Yg"}]}]}} {"custom_id": "Netrakona_16929_29754_17442_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8T2ZhxWwBApejHGTeZnXNz"}]}]}} {"custom_id": "Noakhali_8721_49248_9234_49761.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JE3nxPyiYUFJMBxzP42pcN"}]}]}} {"custom_id": "Sherpur_6156_18981_6669_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GQ6Nr7Eb17iA5PxPa6qVvu"}]}]}} {"custom_id": "Jhenaidah_16929_21546_17442_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X2AgYDnSawrGSZwiNaXQpN"}]}]}} {"custom_id": "Satkhira_10773_5643_11286_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NXtZg5xYF7zhjksxZXsRDi"}]}]}} {"custom_id": "Feni_4617_13338_5130_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JSdWEoVm9B8AQix6gEyQoP"}]}]}} {"custom_id": "Rangamati_13851_7182_14364_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HYjJ9utAD4tSeCuZD9JMCL"}]}]}} {"custom_id": "Mymensingh_7695_40014_8208_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P1PnnMhuc4HQdpErbAegjS"}]}]}} {"custom_id": "Kushtia_2052_11799_2565_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4xExuY9g2taeLRiLjMC8Eo"}]}]}} {"custom_id": "Satkhira_9234_36423_9747_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PLmcmwHKPPEZyBu4W372EF"}]}]}} {"custom_id": "Rajshahi_13338_9747_13851_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JdTme2tK178hhLtGL9vKxt"}]}]}} {"custom_id": "Rangamati_24624_47709_25137_48222.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WkaWGAbXqeog9iHvw4LsTC"}]}]}} {"custom_id": "Pabna_23598_12825_24111_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VpHBZQTJqC96xRRrJmDh9L"}]}]}} {"custom_id": "Lalmonirhat_1026_4104_1539_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Qwnoj4YexCuY94rBSHR5Rq"}]}]}} {"custom_id": "Rangamati_2565_46683_3078_47196.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8ZqxuhUjTUHGakymQCXFaF"}]}]}} {"custom_id": "Habiganj_18981_23085_19494_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VCytghSAtmeenHvrQ93b7N"}]}]}} {"custom_id": "Sirajganj_23598_33345_24111_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SzS3s5f5cfCvxvLTB1CiYw"}]}]}} {"custom_id": "Bagerhat_8721_48735_9234_49248.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U1aZjXdBhvfsuiGzLjivxC"}]}]}} {"custom_id": "Bandarban_7695_49248_8208_49761.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YTvLFsL2dAaQwbtsHrgPju"}]}]}} {"custom_id": "Sylhet_15903_4617_16416_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JTsASKdoiWRS64bPF61ou4"}]}]}} {"custom_id": "Sunamganj_29754_26163_30267_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2DgVWhM4F1ejEGKDwLberi"}]}]}} {"custom_id": "Sirajganj_20007_4617_20520_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JQTkkqygJ2EyM5sgaqLuuP"}]}]}} {"custom_id": "Satkhira_8208_9747_8721_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K4A9vy8tAoqg1aXvKgkWmG"}]}]}} {"custom_id": "Sunamganj_24624_5643_25137_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HrqTFn5j3hVJuiQV9MotMo"}]}]}} {"custom_id": "Jamalpur_3591_32319_4104_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4NJG6zZ9SxkxA8pVBHA7Y5"}]}]}} {"custom_id": "Thakurgaon_17442_12825_17955_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2B4A8VBc41GV6bxjZYnprB"}]}]}} {"custom_id": "Narail_4104_4617_4617_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-92DkS8VmYJCAAiGfGJV6RT"}]}]}} {"custom_id": "Habiganj_5643_16416_6156_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P36d3BTj3AcF3zzFk8pXkT"}]}]}} {"custom_id": "Natore_5643_24624_6156_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YVTQpHydJz7tMtKPXTW42m"}]}]}} {"custom_id": "Khulna_12312_3591_12825_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BnWYtnNNRwh5ASKbykjPrV"}]}]}} {"custom_id": "Bandarban_18981_26163_19494_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LEfjvmAeQU1knhXyVAspDu"}]}]}} {"custom_id": "Noakhali_17955_18468_18468_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SWFLg17WCrQNAZLuoyVRD7"}]}]}} {"custom_id": "Sunamganj_11286_0_11799_513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J1VyPpkVkYrL2i2YbSDdhM"}]}]}} {"custom_id": "Narsingdi_6156_2052_6669_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1fErxNa4mPJ3iR2xpLffZB"}]}]}} {"custom_id": "Bagerhat_14364_10260_14877_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F7fpamo1xkT5oN4eY5X5ZV"}]}]}} {"custom_id": "Kurigram_13338_36423_13851_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L2WSeKPLMXbBSS28VgehuQ"}]}]}} {"custom_id": "Narsingdi_7695_23085_8208_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DCbuMgLrsq1dGEpEf4QiJd"}]}]}} {"custom_id": "Noakhali_4617_1539_5130_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SQC6Q4YJtFj6uejWhZhibd"}]}]}} {"custom_id": "Kishoreganj_8721_8208_9234_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GDgqwg2ApqnkvBY6F2Cpnm"}]}]}} {"custom_id": "Jhenaidah_11286_21033_11799_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8uADXBm46YHAfhFBFuPXgs"}]}]}} {"custom_id": "Tangail_15390_25137_15903_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HZH1pM8UH8f5surqqAAPbr"}]}]}} {"custom_id": "Maulvibazar_13338_22572_13851_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RczPJqLLSf5VisrND1He3W"}]}]}} {"custom_id": "Bogra_25137_3591_25650_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UAVztfmjYLPTHKKQ6TybQw"}]}]}} {"custom_id": "Noakhali_16929_3591_17442_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LLc8VmHXmYZWCKqNF4zhrt"}]}]}} {"custom_id": "Maulvibazar_24624_17442_25137_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8yNPD4VD3PEtZnPGTyqUrj"}]}]}} {"custom_id": "Rangpur_6156_14877_6669_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TTv6SNFAXvHtrAw754E9tJ"}]}]}} {"custom_id": "Munshiganj_5643_7182_6156_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2PbpuP7aex66Nrere2MTrG"}]}]}} {"custom_id": "Nilphamari_2052_17955_2565_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6x399PpA7BMLtNr1YGKtsr"}]}]}} {"custom_id": "Manikganj_16416_11286_16929_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9gN6pYxF4rK2v32Z6jrNeF"}]}]}} {"custom_id": "Bandarban_14877_2565_15390_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MM6EejzbknCBdysPsT7Znd"}]}]}} {"custom_id": "Khagrachhari_7695_34371_8208_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4mo5tof9JMAkjtjGZB2Sje"}]}]}} {"custom_id": "Patuakhali_21546_23598_22059_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FHT9R3DhcjRf2to3kQixiJ"}]}]}} {"custom_id": "Jhalokati_7695_12825_8208_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PiAaofBBAPPdamdxCYgG92"}]}]}} {"custom_id": "Faridpur_28728_13338_29241_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KQq9R3AgHNAdAptfc8oWvS"}]}]}} {"custom_id": "Maulvibazar_10260_19494_10773_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RBwxaVMue6eHszcTcJSoec"}]}]}} {"custom_id": "Magura_10260_2565_10773_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K2aCrYJXxBqE3VLwTQjF8i"}]}]}} {"custom_id": "Dinajpur_11286_3078_11799_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W7uxyEJhF6RBsHfF2ee4Bq"}]}]}} {"custom_id": "Sunamganj_33858_12312_34371_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EFCPhoEYGjqbxggmVQG6QE"}]}]}} {"custom_id": "Patuakhali_21546_9234_22059_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VmkznBTCcagGQbmKMY1o7w"}]}]}} {"custom_id": "Pabna_14364_4104_14877_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PoCMR4N9PMYoEazniTsxwz"}]}]}} {"custom_id": "Noakhali_13338_41040_13851_41553.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9Nb1k9P71eRE6VgQmwQWqN"}]}]}} {"custom_id": "Rangamati_17955_40014_18468_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NryAvHRYem79LPtMir9tth"}]}]}} {"custom_id": "Rangamati_10773_70281_11286_70794.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7NWNxScHqjMZmQsR4CyBd7"}]}]}} {"custom_id": "Tangail_8208_21033_8721_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1QrSuVDdg1yuqb5mJexd3u"}]}]}} {"custom_id": "Bagerhat_15390_5643_15903_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6arqRrcsmhqMM6qcQbgrto"}]}]}} {"custom_id": "Mymensingh_17955_17442_18468_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JU9oDh7nwevp948k1R1fUr"}]}]}} {"custom_id": "Narsingdi_11286_10260_11799_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JvyfhszLusqDmgRgjZsU8y"}]}]}} {"custom_id": "Thakurgaon_13851_16416_14364_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1tPzsHkfTRaJH73YEQ3ffm"}]}]}} {"custom_id": "Sunamganj_32832_12825_33345_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ExvncPAmBLRQatXL9AruWZ"}]}]}} {"custom_id": "Rajshahi_26163_30780_26676_31061.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7xWvYKuw4eF9snPeu6VNTL"}]}]}} {"custom_id": "Habiganj_17955_15903_18468_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6TFGcDqUdhUbwGkmzWvHJZ"}]}]}} {"custom_id": "Nawabganj_17955_14364_18468_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2AKpC8d34yFDURkGTsUiDt"}]}]}} {"custom_id": "Magura_13338_14364_13851_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1Ze7KX52kTv6BgMtGjNtd9"}]}]}} {"custom_id": "Noakhali_18981_14877_19494_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GDmrsCqSYUADBDaj7F2AzY"}]}]}} {"custom_id": "Barisal_24111_11286_24624_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MD2GfqoESoa1Mi4Rd951CV"}]}]}} {"custom_id": "Pabna_14364_2565_14877_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GxnqQKKcWDZ3PNeZujktNz"}]}]}} {"custom_id": "Feni_9747_4617_10260_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SsgdzC8uHdntayvRVGUhyJ"}]}]}} {"custom_id": "Sirajganj_22572_21546_23085_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HjUVYZNeh7TXtkdWad8mnT"}]}]}} {"custom_id": "Jamalpur_4104_18981_4617_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CaY5WZnvZxs2UweedyUEHD"}]}]}} {"custom_id": "Sylhet_1539_22059_2052_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5PXaJrWLuT3UhcqYxgUcuH"}]}]}} {"custom_id": "Pabna_24624_9747_25137_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3tt8wXPiwwRcvNJQW6Z4uZ"}]}]}} {"custom_id": "Natore_1539_28728_2052_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HxQtDKJRxzz43ejFtZ3VuP"}]}]}} {"custom_id": "Thakurgaon_8208_10260_8721_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8n85ihrv89JNJkHuMVuw9p"}]}]}} {"custom_id": "Chittagong_16416_11799_16929_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SzgQGNWpqLk4d6QQ3p2rk3"}]}]}} {"custom_id": "Barguna_12312_27189_12825_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S7sgTJTe5CF5NkfNRinP9V"}]}]}} {"custom_id": "Sylhet_4104_10773_4617_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ua7cztKXQZzex45SdrCjXM"}]}]}} {"custom_id": "Tangail_15903_22059_16416_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LpRf6shFLe9CZNUGpbM3ok"}]}]}} {"custom_id": "Narayanganj_3591_9234_4104_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GNMDMUWc7ZWqbtTmnimVdV"}]}]}} {"custom_id": "Rajshahi_14877_4617_15390_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-45pgbT1BpQaHWJ5buSfeCW"}]}]}} {"custom_id": "Bhola_15390_22572_15903_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Tq3BhB7oJGYRrbJ4YcVcR5"}]}]}} {"custom_id": "Munshiganj_10260_5643_10773_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LcvnB9wjPeBGUeTugP8Q2w"}]}]}} {"custom_id": "Mymensingh_16929_38475_17442_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HKi8TAY5wDVniU9hj8nneB"}]}]}} {"custom_id": "Bandarban_17955_36423_18468_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LJJR9UeVrxY8UwRKdxyrHv"}]}]}} {"custom_id": "Brahamanbaria_23598_16416_24111_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vn4dng9mK5qfTKYLqHWeqS"}]}]}} {"custom_id": "Bogra_1026_18981_1539_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6Ftm3HvdJnRoDLPxDrbzMZ"}]}]}} {"custom_id": "Rajbari_10260_10773_10773_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7yR4gViQD665SvH61qgbxD"}]}]}} {"custom_id": "Rangamati_27702_82080_28215_82593.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hpkk4jPhrXLTussiuo4N7t"}]}]}} {"custom_id": "Mymensingh_17442_18981_17955_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EN5fGEwgKQAfb8BwnU5phJ"}]}]}} {"custom_id": "Sirajganj_18468_6156_18981_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MjMjxcHstrUumgY1JVzZ4q"}]}]}} {"custom_id": "Netrakona_28215_28728_28728_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YaeeZ26zRJSEmBfurYn6oc"}]}]}} {"custom_id": "Chittagong_29241_54378_29754_54891.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AcitrvpXrKfPysv9UiYUA1"}]}]}} {"custom_id": "Faridpur_15903_17442_16416_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9XfUbnakaYNAexg2rr1Qkv"}]}]}} {"custom_id": "Brahamanbaria_21033_6156_21546_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6F1xzgEiphFkBZ8TWopc6g"}]}]}} {"custom_id": "Rajshahi_17955_19494_18468_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-THvn6ioxkM9HG4EByQHiHq"}]}]}} {"custom_id": "Brahamanbaria_16416_18981_16929_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8FT3BuAktMCdH8iRac8e9T"}]}]}} {"custom_id": "Chittagong_26676_24624_27189_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2vtjUBVYdoAAnb7kbAswoT"}]}]}} {"custom_id": "Kishoreganj_30780_13338_31293_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NNVMKxY8Jqj16McWz7M9Qe"}]}]}} {"custom_id": "Thakurgaon_20007_7182_20520_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ga9AmBn4sFgk7AvD1PMvzZ"}]}]}} {"custom_id": "Gopalganj_3078_10260_3591_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WsQqqTbKidq5xDxPZ6ELCk"}]}]}} {"custom_id": "Kishoreganj_26676_9234_27189_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-29BwoqtagFBNz5nGQY5vvC"}]}]}} {"custom_id": "Natore_1026_24624_1539_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KKfXQrBRgvP9wkUCdSoiBs"}]}]}} {"custom_id": "Barisal_23085_4104_23598_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3NfK2mnsjiLYtA4wjYi9wd"}]}]}} {"custom_id": "Sunamganj_20007_7695_20520_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Dh3xbtayWPSJKbnm1x5HbQ"}]}]}} {"custom_id": "Comilla_14877_9747_15390_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ABv1E8PFUFqYxJ9cb4uxw1"}]}]}} {"custom_id": "Sunamganj_11799_4104_12312_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UdNFVUiFMmnhL7oYCSCrLf"}]}]}} {"custom_id": "Kishoreganj_30267_3591_30780_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6FLhoJsKU8hhCvK8LRcSgU"}]}]}} {"custom_id": "Chittagong_17442_21033_17955_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6u1ubyKwt7dV4tkNwuLeZJ"}]}]}} {"custom_id": "Netrakona_19494_10260_20007_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DvJbctRjPpS11eUwto5jSK"}]}]}} {"custom_id": "Chittagong_25650_45657_26163_46170.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VspdjsvWaETtTJdNTRqDGz"}]}]}} {"custom_id": "Maulvibazar_13851_34884_14364_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AStUpBhKp35wzq1h1N3xhv"}]}]}} {"custom_id": "Shariatpur_12312_2565_12825_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PvPnjUDYLLGVsCDRovdUZB"}]}]}} {"custom_id": "Naogaon_20007_7182_20520_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PvEcSLWQd9J5tnCz7nWSo7"}]}]}} {"custom_id": "Rangpur_6669_18981_7182_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KNDGkvAXpBGxehpud5BauD"}]}]}} {"custom_id": "Rangpur_13338_15390_13851_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-92hVVxUf7ES2Rby8aJiiaf"}]}]}} {"custom_id": "Jhenaidah_21033_23085_21546_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FZURtw1HagCBzVEFJENKak"}]}]}} {"custom_id": "Chittagong_21546_10260_22059_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GG9VCobnymXtkYnhW6WEJn"}]}]}} {"custom_id": "Comilla_23085_31293_23598_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PUE9WWJwc3GtnZanEHGG7k"}]}]}} {"custom_id": "Rangamati_26163_51300_26676_51813.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Aj8prQwXYgQhUTNZTgkeqk"}]}]}} {"custom_id": "Nawabganj_7695_8721_8208_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DrTupa7E3wanriLnXuRpy7"}]}]}} {"custom_id": "Pirojpur_7695_15390_8208_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FPJb35faYtA7EMUc8pLh6r"}]}]}} {"custom_id": "Jhalokati_4617_15903_5130_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J5dfBF2vsHr815Rbm5PnNB"}]}]}} {"custom_id": "Nawabganj_13851_5643_14364_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CnEmhWmHtUCYvMauRugjeT"}]}]}} {"custom_id": "Bagerhat_13851_24111_14364_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FZ5yDhsjp7CPyivewWPbDH"}]}]}} {"custom_id": "Rangamati_25137_61560_25650_62073.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9SHao6KP1Sj2dukDhVAeaE"}]}]}} {"custom_id": "Maulvibazar_13338_18981_13851_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W8YiMqN61VKtVuAkXqyzVN"}]}]}} {"custom_id": "Thakurgaon_18468_513_18981_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2mWBK9XcDwYDo1omKcCw5n"}]}]}} {"custom_id": "CoxsBazar_14364_21033_14877_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4u6QGFdxgUrxfZV3quK1aN"}]}]}} {"custom_id": "Sirajganj_26676_36423_26928_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K6StQqMiqFKqxbrZchMEXd"}]}]}} {"custom_id": "Tangail_14877_3591_15390_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UWZ78re1aiNjFMtCjrUMMq"}]}]}} {"custom_id": "Chittagong_27702_18468_28215_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-StwadPBFLhC8qr5FYoivFT"}]}]}} {"custom_id": "Faridpur_15390_15390_15903_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-83XpeYQ8kZaA1ULsv7noCu"}]}]}} {"custom_id": "Khulna_10260_18981_10773_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9r1KZhRgMiGTuNNzDG4yUg"}]}]}} {"custom_id": "Kishoreganj_14877_17442_15390_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9in6DrW2DmRdRoh65Nkn9W"}]}]}} {"custom_id": "Sylhet_23085_19494_23598_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3cjfzoA5PF43XeCwVdV3Qx"}]}]}} {"custom_id": "Rajshahi_25137_13851_25650_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HSPW4GgyxLh2BxTfuonz7j"}]}]}} {"custom_id": "Rajshahi_2565_7695_3078_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Q4mZY3icjPLxCC1KPUc4XM"}]}]}} {"custom_id": "Barguna_13851_18468_14364_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1cW5qLz3MEbXUsSnc9dghM"}]}]}} {"custom_id": "Patuakhali_17955_19494_18468_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MECsz9B1YfvDDGPk572K68"}]}]}} {"custom_id": "Netrakona_25137_25137_25650_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LGvLpyVV7BjtHUsRFf3dDe"}]}]}} {"custom_id": "Comilla_7182_6669_7695_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ArQaUto55dvN5e8PEtRMZL"}]}]}} {"custom_id": "Brahamanbaria_24111_22059_24624_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3eApKQu3Q8LmL8GkUJZFfE"}]}]}} {"custom_id": "Nawabganj_4617_16416_5130_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S2CE1fjJDbV2GiNXUMVuhV"}]}]}} {"custom_id": "Tangail_10773_34371_11286_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LUbQBF8JeojMrPa1v5ofBq"}]}]}} {"custom_id": "Rangpur_10773_17442_11286_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VpNSm3LZFPKy2e3CacEjJk"}]}]}} {"custom_id": "Dinajpur_20520_29241_21033_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1gjPCGH3pwebhVyctLiTNB"}]}]}} {"custom_id": "Mymensingh_14364_18981_14877_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3gnCToHUdTbLxVDAJbPpzD"}]}]}} {"custom_id": "Rajshahi_5130_9747_5643_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9i3wek8VdnKHfR3detW3Ch"}]}]}} {"custom_id": "Shariatpur_7695_3591_8208_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X11uCEAcK72qV4hxt4izUu"}]}]}} {"custom_id": "Dinajpur_8721_15903_9234_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DT2u3uKun8WUetYxeGqY6R"}]}]}} {"custom_id": "Kishoreganj_1539_8721_2052_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FaBwWkJ4QRCpCFXcbuYHwS"}]}]}} {"custom_id": "Tangail_16416_28215_16929_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JNDps3WcZ8bNFT2B1zskWu"}]}]}} {"custom_id": "Rajshahi_18468_8208_18981_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A1tWeFAoeHjXUyt6CLXF4w"}]}]}} {"custom_id": "Jhenaidah_18981_22572_19494_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EbrHuFpTp42zMKUvS9UmxB"}]}]}} {"custom_id": "Bandarban_20520_33345_21033_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4F9FFsE1nbziJj8QGxcYni"}]}]}} {"custom_id": "Habiganj_21033_29241_21546_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1UwDGYTQmytrct6U1aRkBH"}]}]}} {"custom_id": "Sylhet_26676_5130_27189_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7pUUDFULDp39i329W311ME"}]}]}} {"custom_id": "Rangamati_15390_10773_15903_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FwqQiFnFAU617kT7vrms9X"}]}]}} {"custom_id": "Mymensingh_16416_27702_16929_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PhavQDP4E6WCDMxhA3zaGv"}]}]}} {"custom_id": "Tangail_12825_37449_13338_37962.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-79xhD1WbxGUYYBAVZizu3T"}]}]}} {"custom_id": "Sunamganj_30267_18468_30780_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X7L5JouAMUkzHcG98RShUS"}]}]}} {"custom_id": "Sunamganj_9747_23085_10260_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-65LbgavG3yTYpU5tf9CQdU"}]}]}} {"custom_id": "Netrakona_10773_14877_11286_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SxpARX5LtDPxZeeXMS7Qkk"}]}]}} {"custom_id": "Bagerhat_8208_19494_8721_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6yHNepG61odFDvExUyJ8W6"}]}]}} {"custom_id": "Chandpur_4617_24111_5130_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AG8PFrhmm79NhaspTJeVNj"}]}]}} {"custom_id": "Jhenaidah_28215_12312_28728_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KF5DsujNgUozEU8oWqTrgp"}]}]}} {"custom_id": "Bagerhat_18981_20007_19494_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-36px29NTTf3LDiHU2pdkxa"}]}]}} {"custom_id": "Satkhira_12312_18468_12825_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CQ8rSzTVWf41nXH2ZzzpZg"}]}]}} {"custom_id": "Barguna_10260_22572_10773_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QwUrnR8oV8ErmsNYGcQ2Rj"}]}]}} {"custom_id": "Lalmonirhat_17442_22572_17955_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-64W84Wgv9rwr3BgnSwVZUu"}]}]}} {"custom_id": "Patuakhali_14877_35397_15390_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Lq2dBk2qcYbqdUsBMknatV"}]}]}} {"custom_id": "Sunamganj_31806_5130_32319_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CXYPcad8iFCe1HJPsCwHVR"}]}]}} {"custom_id": "Gazipur_5130_15390_5643_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5QSyFY8rEbUHuz9D4fDRqZ"}]}]}} {"custom_id": "Habiganj_5643_13338_6156_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2xpui8DvGUxcD9EvfMBeSp"}]}]}} {"custom_id": "Gaibandha_21033_6156_21546_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-22C3VdbUXhZHn6payfQzJX"}]}]}} {"custom_id": "Narail_9234_16929_9747_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KMrK95DxtRv6HqGqTx53Zo"}]}]}} {"custom_id": "Jamalpur_8208_34371_8721_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KjoR5imq1fACvQmFDHtHqM"}]}]}} {"custom_id": "Shariatpur_7182_12312_7695_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1a3tWH6H2qza8eyfWQ9bBf"}]}]}} {"custom_id": "Jhenaidah_17955_9747_18468_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S3AxKi11uYgb9q8yojNsXd"}]}]}} {"custom_id": "Naogaon_10773_23085_11286_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N1SG2ToPLYNKphwiq9Wqxk"}]}]}} {"custom_id": "Rangpur_11799_16929_12312_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KJhZWQqvcUgJfag7NH7aC7"}]}]}} {"custom_id": "Faridpur_5643_11286_6156_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LzrqEMNhPiTBBUCdsBzebm"}]}]}} {"custom_id": "Nilphamari_5643_4104_6156_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5vBRYUrkj7oTsSAYmNaYLA"}]}]}} {"custom_id": "Tangail_12312_2565_12825_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5L3kM4PgiKbigeJ1mzNqKB"}]}]}} {"custom_id": "Chittagong_30780_24624_31293_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SGFc5Bj5SDM3d7AqdrqEcD"}]}]}} {"custom_id": "Kurigram_15390_26163_15903_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UrUroXa8WQcavHh4bgdnjY"}]}]}} {"custom_id": "Dinajpur_34884_38988_35397_39501.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BkVNTgXkd9Yk5ycf1UDXfb"}]}]}} {"custom_id": "Sunamganj_12312_5643_12825_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QU1aERogEoZnePzXPi7WdH"}]}]}} {"custom_id": "Patuakhali_19494_11799_20007_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gk2xa5EzuBqgv49CUnkH86"}]}]}} {"custom_id": "Chittagong_29754_34884_30267_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XvGomuvFJFf9yFqznRUZhL"}]}]}} {"custom_id": "Rangamati_27189_58995_27702_59508.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WYe41La7fcMRhyiz87ydZT"}]}]}} {"custom_id": "Shariatpur_5643_3078_6156_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vpf2e7bGEFGSjnZHMtgbHG"}]}]}} {"custom_id": "Bogra_21546_3078_22059_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1iLBjSjhWgSrs3x9UiVZ4h"}]}]}} {"custom_id": "Tangail_1026_13851_1539_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DQUPT9rzShnx5krrm5S5Y8"}]}]}} {"custom_id": "Gopalganj_11286_15903_11799_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PaVkHueg9XQuawhNiWitcJ"}]}]}} {"custom_id": "Bhola_4104_7695_4617_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KUFN8vPZzJFEEeY56bgEt5"}]}]}} {"custom_id": "Panchagarh_8721_18468_9234_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P4tU2sPyR395wPKiS8xYJ5"}]}]}} {"custom_id": "Mymensingh_11286_513_11799_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FaM58ycFbgCcrUCtWsjXS8"}]}]}} {"custom_id": "Naogaon_2565_13338_3078_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M6zaJZ1GNmrBsxJ7M1AWAg"}]}]}} {"custom_id": "Sylhet_7695_18468_8208_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NaNwDvXPeaqYL5eHxriqKy"}]}]}} {"custom_id": "Jamalpur_8721_6669_9234_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-H5tY5GZFbWvQHrJ6dJEu8i"}]}]}} {"custom_id": "Nawabganj_6669_14364_7182_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W3h4pnXCmjY5Y1ym5SvCSG"}]}]}} {"custom_id": "Rangamati_4617_49761_5130_50274.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2sdujkhg1FpBh68kBSWNew"}]}]}} {"custom_id": "Munshiganj_15390_6156_15903_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-92R1eu16VzG9kGw3ZueS5r"}]}]}} {"custom_id": "Nilphamari_13338_15903_13851_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GdwbnhTGUsCoxVt3oFJ6vq"}]}]}} {"custom_id": "Dinajpur_26163_24624_26676_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1xmaKoeJPAmqPQTtM5bHPn"}]}]}} {"custom_id": "Chuadanga_6669_14877_7182_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HfTAHzVs8XJ6osAf62tRKE"}]}]}} {"custom_id": "Bagerhat_13851_4104_14364_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AdF5rAzbkSqJXrhRFq9AgL"}]}]}} {"custom_id": "Dinajpur_26163_22572_26676_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RhcoQMagdZ92B39SJkdLYu"}]}]}} {"custom_id": "Bandarban_22572_32319_23085_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3p4oW63ceWXsMH45PyH6UC"}]}]}} {"custom_id": "Tangail_13851_23085_14364_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UhZysvrpDjDG46gUttpg3L"}]}]}} {"custom_id": "CoxsBazar_10260_9747_10773_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XzoQq9R6PCxdatsqoj9LHM"}]}]}} {"custom_id": "Joypurhat_11799_12312_12312_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XFSebsPFQy6zi6YttRxFFz"}]}]}} {"custom_id": "Nilphamari_8208_23085_8721_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EWspesveucxqYHMkGewwNc"}]}]}} {"custom_id": "Naogaon_20007_15390_20520_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K1FYpkdWhLmCMcoxa66pn5"}]}]}} {"custom_id": "Mymensingh_6669_32319_7182_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Rz1pwoNocnH7V6ULzVNtg7"}]}]}} {"custom_id": "Chandpur_8721_13338_9234_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5yw9tm1p83XR4HoTtqivRp"}]}]}} {"custom_id": "Rajshahi_17442_11286_17955_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ex5FiGEMpzy6WRWRjJ8S8v"}]}]}} {"custom_id": "Feni_11286_6669_11799_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8VxsRdY2jz6SPN8ePN4CET"}]}]}} {"custom_id": "Khulna_16929_30267_17442_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-24cmGNhJJkC7X5sEeSKmXK"}]}]}} {"custom_id": "Jamalpur_17442_28215_17955_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8jETeCzUPJEhnhPUfEDCB9"}]}]}} {"custom_id": "Kishoreganj_20007_13851_20520_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TzfiR41Rv1mi21WyUe4Rha"}]}]}} {"custom_id": "Rajshahi_22059_11286_22572_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9J1FuCznGFCbUwJPoMmLQR"}]}]}} {"custom_id": "Khulna_7182_29754_7695_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-H8E2rr3Ep7oWXZgiWh7Vwr"}]}]}} {"custom_id": "Mymensingh_29241_26163_29754_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JaqWCrXFXrfaVgARJDBeLg"}]}]}} {"custom_id": "Pabna_18468_8208_18981_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-21x5Qg7MmxGVu7cvH5Z8YL"}]}]}} {"custom_id": "Narayanganj_4104_5643_4617_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4EzaiRTX9hJihSQbkoTdj6"}]}]}} {"custom_id": "Rangamati_13338_27189_13851_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UcDLg4c9tAZrTabmSJpkFC"}]}]}} {"custom_id": "Chuadanga_12312_19494_12825_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LZyNYgYvHVxXmwjZuQ9h2X"}]}]}} {"custom_id": "Sunamganj_5130_11799_5643_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BXvJoiKbq5AkwqX5bcWWtk"}]}]}} {"custom_id": "Narail_12825_16929_13338_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-17iwV6nq93g3uhiVDje8Bk"}]}]}} {"custom_id": "Satkhira_15390_33858_15903_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VS4RfNWsC7iRmJru2m8c7e"}]}]}} {"custom_id": "Jhenaidah_7182_27189_7695_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ep9adWuWgt7hfCp2eFwbwi"}]}]}} {"custom_id": "Sherpur_3591_16929_4104_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UsEKntCLcWWT8qBVMbE25k"}]}]}} {"custom_id": "Narayanganj_11799_10773_12312_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ly87xE2dRpjEqT9x72w8nf"}]}]}} {"custom_id": "Lakshmipur_12825_26676_13338_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EEHARBmpdJ4jarMpqFLd3j"}]}]}} {"custom_id": "Munshiganj_5643_1026_6156_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NrS1su88wowCkHc8t3HKVx"}]}]}} {"custom_id": "Rajbari_2052_6156_2565_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4oWSMg9Gky7aR3MBw62ZVA"}]}]}} {"custom_id": "Narayanganj_2565_14364_3078_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HU1SkWGUHN4ctLdeov4FQc"}]}]}} {"custom_id": "Habiganj_8721_16416_9234_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3v6W6PSm4uo5VDQ6AdiDZB"}]}]}} {"custom_id": "Feni_7182_18981_7695_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8nVbWRnMBMXZ4tHDjSLnRn"}]}]}} {"custom_id": "Gazipur_11799_13851_12312_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1xmtuRWQ3J5qYfvXiEn2uV"}]}]}} {"custom_id": "Bagerhat_13338_42579_13851_43092.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-H4VUUVF3vCfcayNhNBXZ4U"}]}]}} {"custom_id": "Sylhet_11799_9234_12312_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8rLkLrDc5R28revkBksBkp"}]}]}} {"custom_id": "Bogra_11799_27702_12312_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1qbkUrLoo6kQcoAuL3jX4H"}]}]}} {"custom_id": "Gazipur_15903_6669_16416_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TiQH6a7BTYbtNSK5h9HXxh"}]}]}} {"custom_id": "Chittagong_9234_29241_9747_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1ZvFvy3RN5YEiJpEbb4STf"}]}]}} {"custom_id": "Rajshahi_10260_12825_10773_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vb6mpAH4L8YhQduxPwsWmz"}]}]}} {"custom_id": "Comilla_11286_17955_11799_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MbqXJ3ihJhqcLAqwZ2GBte"}]}]}} {"custom_id": "Sunamganj_6669_513_7182_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ELVpPCoVzPGchSntaRriMj"}]}]}} {"custom_id": "Lakshmipur_3078_8721_3591_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5wpWtPvdkdGRyr6x8FbuJG"}]}]}} {"custom_id": "Sirajganj_10260_14877_10773_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MeqvFjhocw1z9YXPRsQzCy"}]}]}} {"custom_id": "Naogaon_15903_24624_16416_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LVhBGKjCbs3PxPaEJpdCEe"}]}]}} {"custom_id": "Dinajpur_37962_37449_38475_37962.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F29sqt3TP4JHWZkVQkc2yc"}]}]}} {"custom_id": "Sunamganj_11286_23085_11799_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UoyXf6UJQmjoNU86nUxJjz"}]}]}} {"custom_id": "Jessore_17442_14364_17955_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UywYrAeGXX8sh4scLamXUM"}]}]}} {"custom_id": "Khulna_6156_28728_6669_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RYJZPhrDq5yWKQuUEMm5py"}]}]}} {"custom_id": "Natore_5643_10773_6156_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7XpWXs5hyN8SWxCkc83zVR"}]}]}} {"custom_id": "Khulna_8721_21033_9234_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8QwXk7o1ueahfWnauWP61g"}]}]}} {"custom_id": "Mymensingh_15390_6669_15903_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HXiGS9fvGZPRDXGhCPN5xH"}]}]}} {"custom_id": "Kishoreganj_22059_8721_22572_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GZFH9gXwWSojWBrU1jmqp9"}]}]}} {"custom_id": "Chittagong_14877_7182_15390_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KkGBNLRHVogB4Hm62nCww3"}]}]}} {"custom_id": "Rangamati_11799_65664_12312_66177.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P6LBdV1tWS6jPCfWUAkLHn"}]}]}} {"custom_id": "Barisal_5130_4617_5643_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QigzbyoF8yHaKzaCesZZsH"}]}]}} {"custom_id": "Sirajganj_22572_23598_23085_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QEk8r4SEBff6uW49xfRm3Q"}]}]}} {"custom_id": "Khagrachhari_15390_4617_15903_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FcMVzysAyHNcCb8RsjcCri"}]}]}} {"custom_id": "Chittagong_26163_28215_26676_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WWHo3K3Hjn47AjGcym9twL"}]}]}} {"custom_id": "Bagerhat_15390_32319_15903_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TkaUNkixfueRkkw7hJoJbF"}]}]}} {"custom_id": "Rangpur_7182_14364_7695_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3RWQ3QquQqArYBhLHjVsnC"}]}]}} {"custom_id": "Bandarban_11286_20520_11799_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MmiBEai9yXQuKVFJPn5x93"}]}]}} {"custom_id": "Noakhali_5130_52839_5643_53352.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1vM194tnyNRjjGUNgtzhzz"}]}]}} {"custom_id": "Bogra_8721_9747_9234_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6LsSCny8kRg4r4xxH45mKU"}]}]}} {"custom_id": "Dinajpur_20007_13851_20520_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GwgX1TDFubgt4S1prLkKSg"}]}]}} {"custom_id": "Natore_6669_7182_7182_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-28QcjNTHDdvNsyanGDanyo"}]}]}} {"custom_id": "Madaripur_7695_12825_8208_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GawM71KKFQXBxQ9NYPiVVu"}]}]}} {"custom_id": "Bogra_34884_15903_35397_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-35hK6thJ1SUbJ2rNATTdeh"}]}]}} {"custom_id": "Dinajpur_29754_40527_30267_41040.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F7ATk9dyhAZHmzEQxa24Tn"}]}]}} {"custom_id": "Chittagong_24111_30780_24624_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PgQH4GnsM9uWDdpufzRqk9"}]}]}} {"custom_id": "Kushtia_17442_21033_17955_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-E3kHmwvAZ2xQZHtiP7Tkon"}]}]}} {"custom_id": "Pabna_14877_2052_15390_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LXbjk9YWyHZZ3Hk17LotKt"}]}]}} {"custom_id": "Rajshahi_3591_8721_4104_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K9c12AsPgEAisY8SD4o5Un"}]}]}} {"custom_id": "Barisal_7182_4104_7695_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-E64AVY9sjks6gwiUu3yAvx"}]}]}} {"custom_id": "Dinajpur_15390_7182_15903_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4HQNcBS1YEvijPNywFccGw"}]}]}} {"custom_id": "Gaibandha_16416_11286_16929_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W9vaqMGyV5YyWefJgqEhCF"}]}]}} {"custom_id": "Gopalganj_4617_15390_5130_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WnhYLnjoTzJnWNWZtZN1AF"}]}]}} {"custom_id": "Bogra_21546_21546_22059_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F7LvYV1nYoCeTxioS4C9HT"}]}]}} {"custom_id": "Thakurgaon_18468_3078_18981_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8sLYHmtF1GB4iknWQ5UQJr"}]}]}} {"custom_id": "Dinajpur_22572_29241_23085_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JH1pcPEptPsAbZxHDtaw86"}]}]}} {"custom_id": "Dinajpur_32319_27702_32832_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N8mQkKbpY6ep96tun64hBu"}]}]}} {"custom_id": "Kurigram_16416_34884_16929_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3TZriYc1YQLohoKALqpNA9"}]}]}} {"custom_id": "Tangail_10773_4104_11286_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DRTu5zpWaa2DLEM4hPQmKR"}]}]}} {"custom_id": "Gaibandha_17442_20520_17955_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TmfiStFeDWAjzcvchXuQR4"}]}]}} {"custom_id": "Dinajpur_16416_24624_16929_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KuMkob9fZwMBnUktmSHknq"}]}]}} {"custom_id": "Dinajpur_23598_30267_24111_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KVuBbwdYNm8Mvv3WeEygpM"}]}]}} {"custom_id": "Natore_3591_28215_4104_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WfzChDted9TnmvXLqbGvBM"}]}]}} {"custom_id": "Comilla_13338_4104_13851_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UcRXUAS9Kt8JyXKM87VoXW"}]}]}} {"custom_id": "Nilphamari_8721_24111_9234_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UsCWGZtsjx2ewpFnQhh3Yp"}]}]}} {"custom_id": "Rajshahi_513_15903_1026_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QfUrLnZKicqYaiLeZYinQH"}]}]}} {"custom_id": "Jamalpur_6669_26676_7182_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-C2DDc75kyDUa7FN7n3EsPC"}]}]}} {"custom_id": "Bagerhat_10773_18981_11286_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MoabCASJSUAoMKobbpAVq2"}]}]}} {"custom_id": "Lalmonirhat_12312_22059_12825_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JqC3Pp5jo4tPj7dtSjrS4H"}]}]}} {"custom_id": "Gopalganj_12825_10260_13338_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-G4BiP1TTWSSWnYZCbxB2L4"}]}]}} {"custom_id": "Patuakhali_11286_35397_11799_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Md92ornpuApXfHJFkKDtMq"}]}]}} {"custom_id": "CoxsBazar_2565_18981_3078_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A4M7MPRW4Dvk9o1QgnuFC2"}]}]}} {"custom_id": "Comilla_9747_5643_10260_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FR6Eu1d4WXR2hUt3KCkXoC"}]}]}} {"custom_id": "Chandpur_18981_12312_19494_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WJ7u2aXJw27AgT7fp2mNuS"}]}]}} {"custom_id": "Kurigram_10773_17955_11286_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FpFdaFciigU9LusGWTKiCw"}]}]}} {"custom_id": "Brahamanbaria_12312_16929_12825_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UibjJZ1ejFK48LXk7KzBNB"}]}]}} {"custom_id": "Rangamati_16416_14364_16929_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LELRNXhEYvCm95TowfwMpn"}]}]}} {"custom_id": "Joypurhat_7182_17955_7695_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3BJeuEatJDBHGmFucbaFKS"}]}]}} {"custom_id": "Dinajpur_38475_42579_38988_43092.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Jf5c9yw1n9ZBKtWE2btDGX"}]}]}} {"custom_id": "Lakshmipur_15903_12825_16416_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-26nBBbzP6NckvTC1EnMtVs"}]}]}} {"custom_id": "Bhola_7182_44118_7695_44631.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MFnSC6GxC7Uo5DaMPBcVAc"}]}]}} {"custom_id": "Jhalokati_11286_3078_11799_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SHx6kHrAxLkZWx6nKsU22x"}]}]}} {"custom_id": "Feni_5130_23085_5643_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DVeAa9gctVjzngSkMnGdQg"}]}]}} {"custom_id": "Narsingdi_12312_16929_12825_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1KtHDmiDm6WCcQweewLBeg"}]}]}} {"custom_id": "Jhalokati_7182_10260_7695_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6yqfGsJQntockC2wAxA95z"}]}]}} {"custom_id": "Mymensingh_22572_6156_23085_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LHbVDiWXyy4B4zhp9tBqLm"}]}]}} {"custom_id": "Khagrachhari_5643_32832_6156_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ByAcZMZBQk4aD3xCUpU9Ux"}]}]}} {"custom_id": "Gazipur_20520_12312_21033_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DUZ7Y8XQs5VgY4DUW2X5C1"}]}]}} {"custom_id": "Shariatpur_17442_10773_17955_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EcLBW2tXV7zqLDZ1XzqksC"}]}]}} {"custom_id": "Gazipur_17955_9747_18468_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7eo4atzjFxwxnMRZoCHMNw"}]}]}} {"custom_id": "Natore_1539_25650_2052_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-E1FzAKWXff3Kn4MWrQv2hW"}]}]}} {"custom_id": "Madaripur_13338_24111_13851_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5RZhwvw3NQGRayRKJ4YK8H"}]}]}} {"custom_id": "Khagrachhari_7182_48735_7695_49248.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2NGJ4Hp91jw8uYp9bpsxtv"}]}]}} {"custom_id": "Netrakona_20007_12312_20520_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WbnyZp2Y8ZqdLLqTSEEqBn"}]}]}} {"custom_id": "Patuakhali_14877_17955_15390_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SDxB5rkBtEBRNhc6aXAGZS"}]}]}} {"custom_id": "Naogaon_26163_27189_26676_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5wjRQ3QDyChgnsoDNFLXUK"}]}]}} {"custom_id": "Barguna_10260_7182_10773_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2R6a5QpPMhAtCWgYYUxdFx"}]}]}} {"custom_id": "Lalmonirhat_1539_6156_2052_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Si7gjanuHZRU8RQNFJs34V"}]}]}} {"custom_id": "Tangail_11799_27189_12312_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8M2FhmfJG689A3hZXcRi2h"}]}]}} {"custom_id": "Faridpur_9234_11799_9747_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Lsr1bubyR9FNeqHu1uxWBf"}]}]}} {"custom_id": "Mymensingh_20520_10773_21033_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SuT5ZdAAX5SjGrv6Q9iLX3"}]}]}} {"custom_id": "Lakshmipur_2052_13851_2565_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4tPk3rdB4yMzDtMpXMW27T"}]}]}} {"custom_id": "Brahamanbaria_15903_8721_16416_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6hGQmdvSvvdfk634icYtqb"}]}]}} {"custom_id": "Joypurhat_5643_18468_6156_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VeEyPY94R7p7nL4SEX3kyp"}]}]}} {"custom_id": "Panchagarh_8721_13338_9234_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5EWm5yDARkJTnjZzZ2SNVi"}]}]}} {"custom_id": "Kishoreganj_16416_15390_16929_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U43gxeYC2RJa39mo4sBKfv"}]}]}} {"custom_id": "Rangamati_13851_41553_14364_42066.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QUMNtupgKpT1sQy6CrERZb"}]}]}} {"custom_id": "Khulna_6669_5643_7182_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YUyJrG4Gc23c4MVaVHpJd1"}]}]}} {"custom_id": "Sunamganj_29754_11286_30267_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UnkY3kfGiD2qt9MLs6DmTJ"}]}]}} {"custom_id": "Khulna_14364_25650_14877_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MNtR7o8uUNyeqKMDmLPCaj"}]}]}} {"custom_id": "Sirajganj_5643_19494_6156_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RcdvW7TKcCUgXAx9yUYnLZ"}]}]}} {"custom_id": "Nawabganj_3078_20007_3591_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GHp6CXtX7rcBYtuEFmKBXT"}]}]}} {"custom_id": "Dinajpur_5643_25650_6156_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-URe5yK7wwVSLFtmX7jAu3N"}]}]}} {"custom_id": "Jessore_13338_25137_13851_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DB4LEUGBV27AFiZNx9RcDN"}]}]}} {"custom_id": "Khulna_14877_9234_15390_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6H6LfUX33DHEgk942Hvxmn"}]}]}} {"custom_id": "Sunamganj_36423_5130_36936_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3jkhGXveWJQyBCwPJ6zTk3"}]}]}} {"custom_id": "Sirajganj_21033_28215_21546_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DXpRrxLuvrmcZ4oezKe91x"}]}]}} {"custom_id": "Jhenaidah_14364_11286_14877_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DpXPUcMvZEKFNgvAoBiZKU"}]}]}} {"custom_id": "Bandarban_5643_25137_6156_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9h5B85m1mSM9h5ACaegt6y"}]}]}} {"custom_id": "Bagerhat_18468_21033_18981_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P54FAS2YhYoNDJZwevigPY"}]}]}} {"custom_id": "Kurigram_8208_12312_8721_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1pU6uF2F5Q3wq9kwQDftJG"}]}]}} {"custom_id": "Sunamganj_14364_3078_14877_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SHhUm3yBeEBsn3e7PyfFKr"}]}]}} {"custom_id": "Kishoreganj_17955_12825_18468_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9XeK8Z64AJ2a6xnXvYQ1WP"}]}]}} {"custom_id": "Rajshahi_28215_11286_28728_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ScGecY92FYaZqaAYZYkZzL"}]}]}} {"custom_id": "Pirojpur_11286_13851_11799_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ls4hpzLiKx2E7PTCCru9eB"}]}]}} {"custom_id": "Maulvibazar_21033_20007_21546_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1Caj2LKVqnDzMgCgy5rqFr"}]}]}} {"custom_id": "Joypurhat_11799_7182_12312_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Tbbtm4qBjFYruEqwXXihVx"}]}]}} {"custom_id": "Barguna_10773_8721_11286_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8UUZRY6YgQVtbdAFtCDnvN"}]}]}} {"custom_id": "Khulna_16416_10773_16929_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FL1xZ6Uvv23ZWNmDEHKDzx"}]}]}} {"custom_id": "Sylhet_25137_5130_25650_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XeDRv3MvokKVahAcD84PSp"}]}]}} {"custom_id": "Bogra_21546_19494_22059_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XvCyx2zpea4zbCusa4UDvz"}]}]}} {"custom_id": "Khagrachhari_12825_16416_13338_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VcKzQHzZVj9wy6zkDe3v6k"}]}]}} {"custom_id": "Magura_10260_9747_10773_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CrDLsa58g1XFGoLi76Q9UU"}]}]}} {"custom_id": "Habiganj_22059_17955_22572_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bh6HhQH3A9cJ3E4DLiuaZy"}]}]}} {"custom_id": "Bandarban_15903_12825_16416_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EXL5eAt9UsX3181hkuroby"}]}]}} {"custom_id": "Khulna_12825_11799_13338_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CeuagdN7dVLKBaB9YctcrJ"}]}]}} {"custom_id": "Mymensingh_18981_12825_19494_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-97huYtzfYFq4iA9yy6i4LT"}]}]}} {"custom_id": "Naogaon_14364_20520_14877_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GkawmowgsFkawTg3LJFXWE"}]}]}} {"custom_id": "Bogra_20520_6156_21033_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LtQPzfpQwMjeioSSJkyhMq"}]}]}} {"custom_id": "Lakshmipur_9234_9234_9747_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Q9T4agKHYLw17yaNNEmgE7"}]}]}} {"custom_id": "Gazipur_6156_17442_6669_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UckvcGrA4zL1FaSyZYoBuj"}]}]}} {"custom_id": "Manikganj_6156_6156_6669_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X4rAC9ayHyXnHqpdtqF7bn"}]}]}} {"custom_id": "Bagerhat_4617_15390_5130_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QhtLycQGjFK7DGLfZ9Dkvy"}]}]}} {"custom_id": "Mymensingh_7695_15390_8208_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VeunvUqT5F7A2wRKmhHPc2"}]}]}} {"custom_id": "Chittagong_13851_9234_14364_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UWB8oNRkZZZmxWLGh2Sobi"}]}]}} {"custom_id": "Gazipur_8721_14364_9234_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QYwCRGYEw379CdhKrRLXBf"}]}]}} {"custom_id": "Khagrachhari_3591_27702_4104_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QZoJ92nfNdr3f99Nm4MfZL"}]}]}} {"custom_id": "Sylhet_21546_1539_22059_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MdGv8bJPeVRXmwtxfxwbn8"}]}]}} {"custom_id": "Comilla_22572_5643_23085_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2NB8pKvCBndJjekcpY7B9G"}]}]}} {"custom_id": "Mymensingh_11799_45144_12312_45657.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RfUsjKP9UqmU62QYrH1Yig"}]}]}} {"custom_id": "Habiganj_21546_21033_22059_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VGwFLjCan74zbwYxphDk5e"}]}]}} {"custom_id": "Kishoreganj_4104_11286_4617_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ph8GCxAxtYfosHBvqNqsbZ"}]}]}} {"custom_id": "Bhola_22572_29241_23085_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2vBsApySNFxAKXBNRXc3gE"}]}]}} {"custom_id": "Bhola_1539_2565_2052_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QKHSPNhvYswHs2GpBxjGCA"}]}]}} {"custom_id": "Bandarban_13338_18981_13851_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6xhctApERyEnfEzWqA4GJo"}]}]}} {"custom_id": "Tangail_15390_37449_15903_37962.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7iqyt6uV12gtaUeXyB5LtZ"}]}]}} {"custom_id": "Khulna_12312_6156_12825_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KU9dNbvfqx1AUPBzBcnT6a"}]}]}} {"custom_id": "Jamalpur_6669_1026_7182_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-31WsKopCGAjZtybUhPWsLj"}]}]}} {"custom_id": "Sirajganj_16416_30780_16929_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MDTzSM7nzY1EHSvSN2FZRK"}]}]}} {"custom_id": "CoxsBazar_7182_6669_7695_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FUxs4MkjQmvfD4y4uxSPX1"}]}]}} {"custom_id": "Chuadanga_12825_6669_13338_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T483xBnvF72JPvtADKkaut"}]}]}} {"custom_id": "Joypurhat_4617_6669_5130_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F22TpuTehbzdmbDbeSKaPh"}]}]}} {"custom_id": "Brahamanbaria_11286_15903_11799_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GJzxEh3Hbxoe8zG3Qd3e3S"}]}]}} {"custom_id": "Barguna_18981_11286_19494_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5srgQnJdcSKw2WiSA8sPfd"}]}]}} {"custom_id": "Kishoreganj_24624_18468_25137_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2LTDjNpNDEhUpoQ8qsaF8u"}]}]}} {"custom_id": "Mymensingh_9234_7695_9747_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JG8WBqAB7UXZ4KBJymHNpP"}]}]}} {"custom_id": "CoxsBazar_7182_3078_7695_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MuGd6Ae2KCQ9Y28AKdgjXS"}]}]}} {"custom_id": "Rajbari_16416_6156_16929_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DEue7ArmRfQGDthRokT41Q"}]}]}} {"custom_id": "Khulna_14364_35910_14877_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6q3qm4KLjE8YUwsv92SrQ8"}]}]}} {"custom_id": "Bogra_23085_17442_23598_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EWcSjCTcpo2AMbqLx8gxLU"}]}]}} {"custom_id": "Dinajpur_8208_9234_8721_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LFKzckYGDD66MnXYPufaLH"}]}]}} {"custom_id": "Netrakona_513_12312_1026_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JGgGCGzQx1ZL6nNSMuVNDd"}]}]}} {"custom_id": "Jamalpur_4617_4104_5130_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6WU8Gs83cCuZcC25Qe44kp"}]}]}} {"custom_id": "Rangamati_4104_56943_4617_57456.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-19PmsXiNKTVAf3uf2xFmn4"}]}]}} {"custom_id": "Rajbari_12312_13338_12825_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JBRSXt72vUYd5TyszBkyKW"}]}]}} {"custom_id": "Netrakona_18468_8721_18981_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DmS55FjeYaWKA2KF5Zsmst"}]}]}} {"custom_id": "Khulna_13338_31806_13851_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DkrMQbR7BADy4zjoZF9wKf"}]}]}} {"custom_id": "Rangamati_5130_46170_5643_46683.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XaLYTmEV868Co3sKGaNDHL"}]}]}} {"custom_id": "Khulna_3591_36936_4104_37449.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FzdEHwTufSiKWyJmmMXZug"}]}]}} {"custom_id": "Khagrachhari_10773_4104_11286_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UFvAxsCdewt6MHGdu1GvJ9"}]}]}} {"custom_id": "Nilphamari_12825_19494_13338_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UuDS2y5yEBwbchQNZxqUxc"}]}]}} {"custom_id": "Feni_8721_12825_9234_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CSvDvHsZVGBCWEhd7zsaut"}]}]}} {"custom_id": "Chuadanga_4617_14364_5130_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AieqAyyF65H2eyGcg8RnJj"}]}]}} {"custom_id": "Bagerhat_12825_53865_13338_54378.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-89ysVwaaxQGR4X9GG193EQ"}]}]}} {"custom_id": "Panchagarh_10260_20520_10773_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8br4z5XXGFdaivQZLhaWew"}]}]}} {"custom_id": "Jhenaidah_25650_2052_26163_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NHGfbBEn2XEMTn8PNAyp7h"}]}]}} {"custom_id": "Dinajpur_18981_21546_19494_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LQ5dHYwswGYQDFiUZzJzNi"}]}]}} {"custom_id": "Comilla_21033_27189_21546_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5cCkYV22e3aPZGhXxPu14G"}]}]}} {"custom_id": "Netrakona_33858_27189_34371_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7KmVTkw9KjJqcqviAW5u5n"}]}]}} {"custom_id": "Joypurhat_11799_10260_12312_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MFhz6XPX6ofhUjdLL8zP7r"}]}]}} {"custom_id": "Chittagong_30267_42066_30780_42579.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UzpbYfFgRTAofvhTgo9aMa"}]}]}} {"custom_id": "Rangpur_10773_22572_11286_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9tDLWDqvAiiBMic8iky9y1"}]}]}} {"custom_id": "Sylhet_1026_27189_1539_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QY4xxzPej2zgpEeXrXXJ8V"}]}]}} {"custom_id": "Kishoreganj_23598_16929_24111_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KCn7Nsimk7haZ4gVyBLQhK"}]}]}} {"custom_id": "Satkhira_14364_47709_14877_48222.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GFBeU6157Bm7Un1rwQq5wo"}]}]}} {"custom_id": "Natore_20007_2565_20520_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RWiMVhtEHbjtXHpLRDrZDr"}]}]}} {"custom_id": "Magura_16416_18468_16929_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6oDAHaMJegJQUkZFjvAEpg"}]}]}} {"custom_id": "Rangpur_513_10260_1026_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3UeB41CC12MVeu32yJ1BLi"}]}]}} {"custom_id": "Satkhira_4104_13338_4617_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6iwTyNHsqafCMndPe8Vv5b"}]}]}} {"custom_id": "Netrakona_6669_513_7182_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5JGY4tdGvWaojTByePBwWX"}]}]}} {"custom_id": "Tangail_18468_16416_18981_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XPFbriHGfWnxiCqdjXwS2x"}]}]}} {"custom_id": "Kushtia_4617_7182_5130_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NAfRSMvCkQqugB7zeAt27y"}]}]}} {"custom_id": "Rangamati_26676_64125_27189_64638.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S5XXst7FkzKeWGuFX8mvqz"}]}]}} {"custom_id": "Jhenaidah_4104_25650_4617_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4Vyny2t8NdQ4yhcf55zwUS"}]}]}} {"custom_id": "Gaibandha_21033_16416_21546_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BH7Xz2mxHZCYLGDhBAuEZB"}]}]}} {"custom_id": "Thakurgaon_12312_7182_12825_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BcCcYFihXnSnDP4uG4szYs"}]}]}} {"custom_id": "Comilla_17955_14364_18468_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bxk8anxqdUEnMAoMGKJtUK"}]}]}} {"custom_id": "Narail_5130_4617_5643_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8PSvBGjkZBJdJG9VrZqqfg"}]}]}} {"custom_id": "Chittagong_35910_52839_36423_53352.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VmwCyLTi9howty1Q5jaCx7"}]}]}} {"custom_id": "Chandpur_13851_3078_14364_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2jYKaNRh5Z8NjWkaREDv2Y"}]}]}} {"custom_id": "Sylhet_2052_29241_2565_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BNRE8CENfQCxh8oarFfDhr"}]}]}} {"custom_id": "Sunamganj_12825_10260_13338_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KiS312HtCZjuKoQQijnfkd"}]}]}} {"custom_id": "Gaibandha_9234_11799_9747_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NESgVTY7EJVmMXLau8kM5C"}]}]}} {"custom_id": "Gopalganj_10773_12312_11286_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VioqnhfkbHUQzyn2GCDNPJ"}]}]}} {"custom_id": "Munshiganj_19494_7695_20007_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3DsbMh8MRBjGs8315dUZMN"}]}]}} {"custom_id": "Khulna_10773_28215_11286_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PsWukAGttS1Zij7ubdScvi"}]}]}} {"custom_id": "Rangamati_24624_50787_25137_51300.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VKuMJgMc4UT7DkcnzDhybK"}]}]}} {"custom_id": "Bagerhat_12312_9747_12825_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FGdBCZZrqRaofEuRtHeDxz"}]}]}} {"custom_id": "Narsingdi_10260_3078_10773_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7nz5CUFVHirSJSd3hPUrcj"}]}]}} {"custom_id": "Patuakhali_12825_16929_13338_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Qryhddhzq83Ch18jE2TeG7"}]}]}} {"custom_id": "Bogra_14877_14877_15390_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QEAzYFbpZQLSR9iTUtLDtk"}]}]}} {"custom_id": "Satkhira_14877_54378_15390_54891.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Y9xFV2eZYZqmS3eLCZK2Fb"}]}]}} {"custom_id": "Natore_15903_6156_16416_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-153f8BSCw5ijCynkQ6GSg3"}]}]}} {"custom_id": "Mymensingh_9234_14877_9747_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NaaKicSjjRz6KEHzj6pfjM"}]}]}} {"custom_id": "Bandarban_15903_4617_16416_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MpEHBWWj5XMcc6Pztj9DCq"}]}]}} {"custom_id": "CoxsBazar_3078_9747_3591_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QCiVesVJzy2MT76Kfaa7j1"}]}]}} {"custom_id": "Sherpur_4617_1026_5130_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8mjd3S68wAcKijHMuSe848"}]}]}} {"custom_id": "Barisal_13851_13338_14364_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6cynATv2ZjdERfrKUou5wL"}]}]}} {"custom_id": "Sylhet_21546_6156_22059_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YJPuJhNTjAkUrnXEs8izVp"}]}]}} {"custom_id": "Barisal_22059_5130_22572_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TA9ywG6aNx3nDXkjZoSfez"}]}]}} {"custom_id": "Narsingdi_9234_22059_9747_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L7DBMQThiLAztoXoudk3Aq"}]}]}} {"custom_id": "Patuakhali_18981_19494_19494_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-77xSAaf3h48K3hzHab49XE"}]}]}} {"custom_id": "Sirajganj_12825_24111_13338_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DUiATx24m97LiBSwCX5MDX"}]}]}} {"custom_id": "Rangpur_9747_14877_10260_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5aFqFjxX9ELVnv488KkgsT"}]}]}} {"custom_id": "CoxsBazar_13851_6156_14364_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LELXDmo4AB3qjuRJuEvF9p"}]}]}} {"custom_id": "Joypurhat_4104_3078_4617_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L5Eb3j71jNhVu7qt4PwvHM"}]}]}} {"custom_id": "Sylhet_13851_24111_14364_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ScU64tgtWRoGzZ6Er2eYe3"}]}]}} {"custom_id": "Rajbari_20007_16416_20520_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W818wj6j8FgX6mAatYao5B"}]}]}} {"custom_id": "Khulna_10260_11286_10773_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KAG67zUovZXmSWk7QBUTM8"}]}]}} {"custom_id": "Netrakona_26676_24111_27189_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NoEmu7ienWcGauc4FEviuz"}]}]}} {"custom_id": "Kushtia_20007_12312_20520_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YHNwtpWRdjuHXP62vcpir3"}]}]}} {"custom_id": "Comilla_26163_28728_26676_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UxNtp271Rpma2ho5enKpCS"}]}]}} {"custom_id": "Rajshahi_27189_6156_27702_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-383GieHkdJA1Srpz6Cg73c"}]}]}} {"custom_id": "Khagrachhari_9234_49248_9747_49761.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DwyKbPs1HN3M7eXzzJBhHx"}]}]}} {"custom_id": "Satkhira_4617_5643_5130_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K8mwvm9UV3TCUU2HScd5K7"}]}]}} {"custom_id": "Khulna_6669_13338_7182_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2HiysZthaLA8TGvS9qygDk"}]}]}} {"custom_id": "Khulna_18468_7695_18981_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R4xURRenG5WLnuwnVQ5rNk"}]}]}} {"custom_id": "Rangamati_4104_51300_4617_51813.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4KktSyhnixx6dr2QMjx9U4"}]}]}} {"custom_id": "Netrakona_10260_17442_10773_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6yaWQTHzfdMoMbDpBJf6BU"}]}]}} {"custom_id": "Noakhali_12312_41553_12825_42066.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T4dcQ5VYLaaXRpg6xhkJm9"}]}]}} {"custom_id": "Joypurhat_5643_4104_6156_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LEKPBhJTcG8mRnEmez8CQw"}]}]}} {"custom_id": "Chittagong_24111_18468_24624_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3o5qj8A19zDqrtBQyXudB5"}]}]}} {"custom_id": "Dinajpur_17442_28728_17955_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WiVWrzf3M8sUMvP6dQZEfx"}]}]}} {"custom_id": "Naogaon_26163_15903_26676_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4qSD54qFoQ2hVdaaX6ZX41"}]}]}} {"custom_id": "CoxsBazar_15903_16929_16416_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CfL6iK11FkaLzUR5XYVfdh"}]}]}} {"custom_id": "Bhola_7182_43605_7695_44118.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-531PpJ58J3UZ38ikTFjjYD"}]}]}} {"custom_id": "Satkhira_20520_50274_21033_50787.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EFQcysdbmGcqq7E8TVfgQq"}]}]}} {"custom_id": "Jessore_23598_17955_24111_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VN2QWr4rVu7hsZqVz3j7fk"}]}]}} {"custom_id": "Comilla_7182_11286_7695_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S2fnaMuUoHgBeWwyvFqgmK"}]}]}} {"custom_id": "Jhalokati_9234_13338_9747_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4GbgfNektBLocHj9cBq9Ei"}]}]}} {"custom_id": "Habiganj_513_20007_1026_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MdFApHBvyZq8YfR9xcwBQG"}]}]}} {"custom_id": "Thakurgaon_16929_10260_17442_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WCdTkWZMAogqAbMcFbRTQQ"}]}]}} {"custom_id": "CoxsBazar_11286_19494_11799_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BC3WyQP5fe7KKKAx9whvWM"}]}]}} {"custom_id": "Chandpur_9747_13851_10260_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HTrDrrFrYut7NYVG2g6Vzy"}]}]}} {"custom_id": "Tangail_7182_22572_7695_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EnkucTWuCLugCK7kcKjy8L"}]}]}} {"custom_id": "Tangail_21546_31806_22059_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FWhcWN395tDfUssCspX9vs"}]}]}} {"custom_id": "Khulna_3591_27702_4104_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CKZVfYccqMYjJMe7LzUnUR"}]}]}} {"custom_id": "Rangamati_20007_50787_20520_51300.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M2Hhp83LPq4XFeFgMtAEKn"}]}]}} {"custom_id": "Barguna_18468_22059_18981_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X2Aqi45tzGYAgdSVMutyRM"}]}]}} {"custom_id": "Madaripur_10773_11799_11286_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JtmCiyyHgD7KkFnT2R1HA9"}]}]}} {"custom_id": "Maulvibazar_8721_17442_9234_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BvcVFNTrX7VqR4RiUdGgyY"}]}]}} {"custom_id": "Bagerhat_7695_20520_8208_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VaNFwbAf23A3cUGoqFTS55"}]}]}} {"custom_id": "Rangamati_13338_33858_13851_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1UXL7A7G9tLNzJwfjhAAb6"}]}]}} {"custom_id": "Bogra_18468_27702_18981_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KSSr7i9yyRiGoDvd4Bv5cg"}]}]}} {"custom_id": "Natore_21546_17955_22059_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CprEakH8PDumLbY2QCmBdY"}]}]}} {"custom_id": "Maulvibazar_8721_11286_9234_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CaWRMR6TbnKLGEmZHVhnj6"}]}]}} {"custom_id": "Naogaon_16416_25650_16929_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GGSDKacdDduEbj59hXehG1"}]}]}} {"custom_id": "Rajbari_8208_8208_8721_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GZ6rtvH3rTgeenB2hvaSwo"}]}]}} {"custom_id": "Dinajpur_13338_14877_13851_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Kg2ocqvUcTuXN4myssMcTx"}]}]}} {"custom_id": "Rangamati_16929_31293_17442_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GFytHk8vLsscGoUe2pmREa"}]}]}} {"custom_id": "Barisal_20007_9747_20520_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1Bf6xrhBe89BPGFNtAXXpB"}]}]}} {"custom_id": "Mymensingh_7695_23598_8208_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N8dEMLQ69M38f4rrz5pW2H"}]}]}} {"custom_id": "Satkhira_11799_13338_12312_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gn9TyfdJAuEPfTfy8JEwW3"}]}]}} {"custom_id": "Satkhira_9747_18468_10260_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5FKgyv8MFbdHms1ntHcbbi"}]}]}} {"custom_id": "Bandarban_13338_20520_13851_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A9j29N489MoaiqXSh6ZHQK"}]}]}} {"custom_id": "Chandpur_15390_10260_15903_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JHCQPnWwDLQ4tf3AyQ1PLx"}]}]}} {"custom_id": "Habiganj_10260_15903_10773_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DuypnRKWNa8i5SGSVZp6km"}]}]}} {"custom_id": "Jhenaidah_23085_21546_23598_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1a721BNVVssuYeARyYG83s"}]}]}} {"custom_id": "Bogra_27189_9234_27702_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-95Bmpqyi7YJKmEgyvDETgg"}]}]}} {"custom_id": "Manikganj_12312_18981_12825_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NZu1PN1SxWHtb6w1jThDwu"}]}]}} {"custom_id": "Netrakona_6669_8208_7182_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HRYzpKzYAqbDtY6LzouJed"}]}]}} {"custom_id": "Bogra_8208_23085_8721_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NzDUS8e7gHSA4ypn46cUr9"}]}]}} {"custom_id": "Satkhira_6669_51813_7182_52326.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5vWW9RRmFasMW5K3S1L5qb"}]}]}} {"custom_id": "Jamalpur_2565_25137_3078_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MBhAcdkJgUR4LkRu1iLtAg"}]}]}} {"custom_id": "Sherpur_10260_16416_10773_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1Q12gM4cLVaCUM2bXVZftd"}]}]}} {"custom_id": "Pirojpur_1026_31806_1539_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3dS8kT88GuND2wDmq3MgPD"}]}]}} {"custom_id": "Kishoreganj_6156_13851_6669_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QY1Sf7yVXCSjAEUP4t8hP4"}]}]}} {"custom_id": "Natore_1539_16416_2052_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vb9C7NnAtMNxDhLS45h358"}]}]}} {"custom_id": "Jamalpur_9747_16929_10260_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UKiJJurzZM2KpunL7fvtxc"}]}]}} {"custom_id": "Madaripur_15903_15903_16416_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5zajmReMDSA6ECyDLdufgq"}]}]}} {"custom_id": "Jamalpur_8208_6669_8721_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SvEoqyRRmk8hiZ4iuqFXqs"}]}]}} {"custom_id": "Gopalganj_6669_2565_7182_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UyFqQEkiZw9ignzTyZdk8q"}]}]}} {"custom_id": "Khulna_8208_19494_8721_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-THjVBWCxLoVR4GQ3Z63K6P"}]}]}} {"custom_id": "Nawabganj_10260_14364_10773_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TZxxKop3YtQMKV4PFYaAP1"}]}]}} {"custom_id": "Pabna_13338_3078_13851_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FD6Rc4kw26fJWwSeay2E4Q"}]}]}} {"custom_id": "Bogra_21033_17442_21546_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1AScm1Y4oXkPvyFHnaa2WY"}]}]}} {"custom_id": "Mymensingh_30267_34371_30780_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2yUyaSrxfpcbEas9MENAP4"}]}]}} {"custom_id": "Patuakhali_19494_31293_20007_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EBX4SAphcumBr2a3ADwvEu"}]}]}} {"custom_id": "Maulvibazar_27702_3078_28215_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UVs8hdWsPx4egDc7vi13z6"}]}]}} {"custom_id": "Chittagong_31293_31293_31806_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KWyocJMExkFxEKE3HKiAM7"}]}]}} {"custom_id": "Brahamanbaria_20520_24624_21033_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-69RBwSNmZ4trmtUz68aPLc"}]}]}} {"custom_id": "Kurigram_12825_7182_13338_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6hFQWRSWrpKbJA5LQ21p82"}]}]}} {"custom_id": "Barisal_20520_23598_21033_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MzzRbWYwAuiM58fhkQYMQf"}]}]}} {"custom_id": "Narayanganj_513_16929_1026_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Q364ZSWAgQzbwMcUJNiKMm"}]}]}} {"custom_id": "Panchagarh_3078_6669_3591_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mo3dQtHjFgXu2CH5VjLP7d"}]}]}} {"custom_id": "Bhola_7695_6156_8208_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AVPNVTkC5AXKyXCBW7MDXM"}]}]}} {"custom_id": "Brahamanbaria_18468_22059_18981_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HarwQYac99pNeZndPZhyg7"}]}]}} {"custom_id": "Rangamati_15903_55917_16416_56430.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WkADKYZi51zAbqzSrWanLP"}]}]}} {"custom_id": "Jessore_22572_8721_23085_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9hz42ABrJU4dJbHt4opuoK"}]}]}} {"custom_id": "Magura_7695_21033_8208_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EgwRE8yfAbuoDcNJ3ZRcwJ"}]}]}} {"custom_id": "Chittagong_18468_8208_18981_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4LPjS3nTs1LvFHA9aT8kHF"}]}]}} {"custom_id": "Bandarban_6156_12825_6669_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7xQdC9bBCrUBCKDDx1yqxq"}]}]}} {"custom_id": "Narayanganj_2565_11286_3078_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3Vw6AAxuWJFHtL8RAxaAog"}]}]}} {"custom_id": "Madaripur_10260_21033_10773_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J9f4FREXENhdM2bExez3Xz"}]}]}} {"custom_id": "Jessore_3078_19494_3591_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JYYsMX5TLEccRShYD3831v"}]}]}} {"custom_id": "Sunamganj_16929_15903_17442_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FTivE89Wp1LNjdtnys7AEe"}]}]}} {"custom_id": "Satkhira_8721_3591_9234_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XNSKBHLksYKLT5ABa2Qeyb"}]}]}} {"custom_id": "Gopalganj_11799_20520_12312_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SUwCzHi1eXS2xEzrAqaeP2"}]}]}} {"custom_id": "Kushtia_10260_17442_10773_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YLsMhhscfS848ET5mxLPhZ"}]}]}} {"custom_id": "Rangamati_15903_1539_16416_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fi3voS9qs8TJoZYEDTWkzG"}]}]}} {"custom_id": "Jessore_12825_14877_13338_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9NTzkgQD4MgYsxWh7yzQxW"}]}]}} {"custom_id": "Patuakhali_20520_26676_21033_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BvBtB15jnDMxzyo1jqsfVk"}]}]}} {"custom_id": "Gaibandha_16416_5643_16929_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S2NGHpFvWJmc9VNF3JNvbE"}]}]}} {"custom_id": "Kushtia_21546_18981_22059_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LmB7VRh6NgakMgHMAQPZdV"}]}]}} {"custom_id": "Patuakhali_3591_15390_4104_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9Msqe7dGVJU5LL5fHb6YiQ"}]}]}} {"custom_id": "Dinajpur_14364_3591_14877_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2ASmLobnxP4uvHQj4H8VsK"}]}]}} {"custom_id": "Khulna_4104_39501_4617_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8HmVKLDq2eAnGTUjjtQxMb"}]}]}} {"custom_id": "Chandpur_12312_2565_12825_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-97W9Ez35rEZF1ZEzcqbNfp"}]}]}} {"custom_id": "Jamalpur_5643_23085_6156_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WXKtngbeyMjJaFeAYXGsko"}]}]}} {"custom_id": "Nawabganj_10773_12312_11286_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3gJNvLfmqTjTPcPCEqWn2t"}]}]}} {"custom_id": "Jhenaidah_22572_15903_23085_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V1PiGEbS3a7ik1bVnYzE1W"}]}]}} {"custom_id": "Khagrachhari_9234_45144_9747_45657.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HCSZVZA5iM1WQrLbmjBpfh"}]}]}} {"custom_id": "Faridpur_5130_12312_5643_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BT6Hw14zfh2ZsfLpiZAREK"}]}]}} {"custom_id": "Sirajganj_22572_10260_23085_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4uxzzHif31Z4QhjyFzUAFW"}]}]}} {"custom_id": "Sunamganj_16929_5643_17442_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JKKAxec9shzRDvnu9n8NNA"}]}]}} {"custom_id": "Magura_9747_13851_10260_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4mV22YSCU9mZS8biqUMKao"}]}]}} {"custom_id": "Kishoreganj_11799_17442_12312_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JSCG8zDHVvJE64Gx9HYpV3"}]}]}} {"custom_id": "Bhola_7695_6669_8208_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-27eKxEFX5nRs7QhiyNyHvY"}]}]}} {"custom_id": "Kurigram_11799_25650_12312_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LhjBb8D7MjPSMm63Wn39Jm"}]}]}} {"custom_id": "Kishoreganj_13338_27189_13851_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LZimPqUQAXwxeSRVkBnw3v"}]}]}} {"custom_id": "Lakshmipur_5130_13338_5643_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J3vwT1mTy8W29P8T3vozBc"}]}]}} {"custom_id": "Faridpur_10773_8721_11286_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J7XxHAuevPigxZnAVAon21"}]}]}} {"custom_id": "Joypurhat_10260_7695_10773_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UYeXHTUgdAoxPGRygdyoSC"}]}]}} {"custom_id": "Comilla_4617_6156_5130_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-C9mYerarphug5D5RQ7Pc3n"}]}]}} {"custom_id": "Khagrachhari_15903_28215_16416_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SvLCZhqevds7wziD3i9rxm"}]}]}} {"custom_id": "Kurigram_6156_27189_6669_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RLrcFi4APT2EC9P2Bc5DoZ"}]}]}} {"custom_id": "Patuakhali_9234_26163_9747_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FSjEZRfwtTuUXk76Tw9ijm"}]}]}} {"custom_id": "Naogaon_3078_13851_3591_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4GvgexXCUYK2wvRv2kcaMw"}]}]}} {"custom_id": "Narail_11799_17442_12312_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NZFGNsX7YTeRsCVheqQU3R"}]}]}} {"custom_id": "Sunamganj_14877_23085_15390_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DZH5NbUMC17Qbru1Aoswv5"}]}]}} {"custom_id": "Bogra_23598_11286_24111_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6AbkDaboYuy9cDFRq1mSaD"}]}]}} {"custom_id": "Narsingdi_5130_1026_5643_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3T7Fvy6V1y4d42ZxXjZek1"}]}]}} {"custom_id": "Naogaon_7695_22572_8208_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R3fRrn5LLLFzGD7qx94qFf"}]}]}} {"custom_id": "Chittagong_22059_24624_22572_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PMqjDdrBWfvdKGcN2UH7Y7"}]}]}} {"custom_id": "Patuakhali_21546_23085_22059_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VZV5wzKCrEvaevwzYWynXM"}]}]}} {"custom_id": "Natore_9747_11286_10260_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-85QZmEpzy1bz6MmoRL5BA7"}]}]}} {"custom_id": "Satkhira_8208_54378_8721_54891.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ue7iKJ9LpKg8Uc6j9M5ebM"}]}]}} {"custom_id": "Satkhira_9747_23085_10260_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AamoHCi9EGRfbyDdXKcWZe"}]}]}} {"custom_id": "Khagrachhari_5643_22572_6156_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RvXPxTRi1HsKMU9BCmHCZ3"}]}]}} {"custom_id": "Shariatpur_6156_7695_6669_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Kjd1nW48QqoF4FQtWsywVR"}]}]}} {"custom_id": "Sherpur_9747_17442_10260_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8T9sfT7LRQmx7u2GeBUxyG"}]}]}} {"custom_id": "Rangamati_18468_63099_18981_63612.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8w4bmGyk9L61Pup4tDH8VK"}]}]}} {"custom_id": "Jessore_14877_25137_15390_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-97bAfppQsCJj8WygXFcm2A"}]}]}} {"custom_id": "Netrakona_5130_16416_5643_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9H5nHLT6nV8MuGEP4Xu8or"}]}]}} {"custom_id": "Netrakona_15390_23085_15903_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PXqkJ8iCTF3R8fz9JaTQJe"}]}]}} {"custom_id": "Chittagong_22059_17442_22572_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QUcgsZQ4BbKwuNYcxFTaDp"}]}]}} {"custom_id": "Khagrachhari_12825_8208_13338_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Qwf5XCCNgf5H4zZxDLRTXH"}]}]}} {"custom_id": "Pirojpur_7182_31806_7695_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EWAFtZTQYaN25kb1zosB3c"}]}]}} {"custom_id": "Sylhet_32832_10260_33345_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9tuyHJr1Sg2MPmvDATQDp6"}]}]}} {"custom_id": "Maulvibazar_1539_19494_2052_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1voDVLuf73iFrtDTot8G5g"}]}]}} {"custom_id": "Jamalpur_11286_14364_11799_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F4czsSPUefX22UqMoX8Tzu"}]}]}} {"custom_id": "Barguna_9234_25650_9747_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LtzGLSuRUzGQinnZC8LcYi"}]}]}} {"custom_id": "Rangamati_9747_62073_10260_62586.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DqJs2eiRPGa4H9QqfJY6qS"}]}]}} {"custom_id": "CoxsBazar_4617_3591_5130_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8FiARQnrC1AyVChxyS4MV6"}]}]}} {"custom_id": "Narayanganj_10260_6669_10773_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5HXPP5z8Cn53kBNdMBfJBx"}]}]}} {"custom_id": "Gopalganj_8208_20520_8721_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7pH8dmTPJcUWAL1TMBmFt8"}]}]}} {"custom_id": "Pirojpur_2565_7182_3078_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bkb8pBdQMNR6pmyEw4W1We"}]}]}} {"custom_id": "Khagrachhari_9747_19494_10260_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UYvzaH4NDXJVAs6UPn1nt2"}]}]}} {"custom_id": "Habiganj_15390_4617_15903_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-The27G5NSZtJxp5XWaEG2s"}]}]}} {"custom_id": "Barisal_18981_12312_19494_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1iahMH3VwCry2TMi9G7HBA"}]}]}} {"custom_id": "Maulvibazar_21546_6156_22059_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-888kmLDL69VbRFmFfLngf6"}]}]}} {"custom_id": "Jamalpur_7695_8721_8208_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UnKTBYS4nwhSGujxSrZrJX"}]}]}} {"custom_id": "Satkhira_8721_43605_9234_44118.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D3mmwXA2X3VkSiYNX2tG3r"}]}]}} {"custom_id": "Brahamanbaria_22059_5130_22572_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U6w5d9nofjheRZAnCgqywP"}]}]}} {"custom_id": "Rangamati_25137_70794_25650_71307.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LAwuJMLojkeT3NRG97rhPq"}]}]}} {"custom_id": "Bandarban_20520_23085_21033_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8WMhNgH1cYSH5EJPvZFbGP"}]}]}} {"custom_id": "Rajshahi_7182_10773_7695_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SyPd2W5u2fFFE1gJdVBeck"}]}]}} {"custom_id": "Noakhali_8721_45144_9234_45657.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D7A5R9QUGbk39iWYQ4B7of"}]}]}} {"custom_id": "Bandarban_12825_9747_13338_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3WfQWaVVJPqRoEmtoLJTeQ"}]}]}} {"custom_id": "Chittagong_34884_27702_35397_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L9rDVD2jLp72HLEmagZovW"}]}]}} {"custom_id": "Kurigram_6156_13851_6669_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MQfouaptmw3aXa3VKsADJd"}]}]}} {"custom_id": "Pirojpur_5643_7182_6156_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K2mEgKyrjAZgwePDEyjve6"}]}]}} {"custom_id": "Chuadanga_2565_11286_3078_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fb7kgWfsSRXbEykheZXeS6"}]}]}} {"custom_id": "Netrakona_8721_11286_9234_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HQ7qY3xVHQQdokkU1Ra7Ha"}]}]}} {"custom_id": "Pabna_8208_16416_8721_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5aBaPxLUpX1rip9qtDr1Qs"}]}]}} {"custom_id": "Kurigram_15390_43092_15903_43605.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8EMV94yZ8tbm22NMpCM1vs"}]}]}} {"custom_id": "Kushtia_28728_15903_29241_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QxGtm8v27DNtQ5ep8PybNv"}]}]}} {"custom_id": "Bandarban_2052_21546_2565_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YBG1zMbc2JQGe53S8p73R7"}]}]}} {"custom_id": "Tangail_15903_21033_16416_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Tcxb9SxCtsHRFJgjzQ5HW1"}]}]}} {"custom_id": "Bhola_13338_19494_13851_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5NWbYWh3Ba2h9EnShjCfvR"}]}]}} {"custom_id": "Chittagong_25137_12312_25650_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VjswN1e9cDNsxxJYitLBiZ"}]}]}} {"custom_id": "Magura_14877_10773_15390_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4hy1vxowFZUAjsCs9RP2FY"}]}]}} {"custom_id": "Rangamati_18468_64638_18981_65151.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MH7KmTeBjKz2m1TeyC9t5a"}]}]}} {"custom_id": "Manikganj_21033_10773_21546_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EVVHczhtSn8qBnbHREdsNc"}]}]}} {"custom_id": "Natore_7182_13338_7695_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FfxG9nba3LE846CPa4fTMm"}]}]}} {"custom_id": "Lalmonirhat_15903_17955_16416_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GzPbCdwxoYr76SW1RK3ntZ"}]}]}} {"custom_id": "Nilphamari_13851_24111_14364_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AHV5Z44Qb9CWNm8WoZxtbT"}]}]}} {"custom_id": "Rangamati_6156_51813_6669_52326.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TjjzKdB4ZrdP4NCWx62pT7"}]}]}} {"custom_id": "Netrakona_3591_19494_4104_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QJsdYU34gXb8vehuvw9g9P"}]}]}} {"custom_id": "Tangail_10773_25650_11286_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MMmrNXqKB7JU5teFGRa79k"}]}]}} {"custom_id": "Naogaon_23085_8208_23598_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AwNXSSs7Kn6enXthpgbU4E"}]}]}} {"custom_id": "Satkhira_10260_35397_10773_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ScPPEQqR4sx4JSFG1BtR1j"}]}]}} {"custom_id": "Kushtia_9234_9234_9747_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9PWYxpYY4PV6DAf3BfY7qB"}]}]}} {"custom_id": "Sirajganj_18981_30780_19494_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DJ6jESgW2bfbLqdR6VtQmQ"}]}]}} {"custom_id": "Kurigram_5643_15390_6156_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9mkMziHcmbJ9vJ41nwxNpd"}]}]}} {"custom_id": "Joypurhat_9234_10773_9747_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LrfFtYmb8Gkqkj1isGLR5K"}]}]}} {"custom_id": "Khagrachhari_8721_37962_9234_38475.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gs1MnLBjzq3qtL3h4XPNXJ"}]}]}} {"custom_id": "Chittagong_35910_45144_36423_45657.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8xU5cHAdsJTitLBevmfURr"}]}]}} {"custom_id": "Dinajpur_26163_32319_26676_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NmJmNPj8FXNoBuHqmNHFoq"}]}]}} {"custom_id": "Munshiganj_5130_9747_5643_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HntpdyoHfcdbYmr8W9gxP3"}]}]}} {"custom_id": "Meherpur_14877_5643_15333_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HGfxVAyYqfMmCmiZyzuZWz"}]}]}} {"custom_id": "Comilla_31293_35910_31806_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1V9qviZ4absggvRxLdqTLX"}]}]}} {"custom_id": "Bogra_23598_3591_24111_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SqTDyUYRzcT2pwp5znnjkw"}]}]}} {"custom_id": "Rajshahi_23085_14364_23598_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RMmt5fHcVXZ8r39WE3U69B"}]}]}} {"custom_id": "Barguna_14364_5130_14877_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PfPCW9THhiZKvipmBWqh9P"}]}]}} {"custom_id": "Bagerhat_13338_47196_13851_47709.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MEAivSi48GddHkBpFQdv6q"}]}]}} {"custom_id": "Rangamati_6156_48735_6669_49248.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BAKaMtekqh9PvQTJNJDrQE"}]}]}} {"custom_id": "Dinajpur_1539_15390_2052_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LjNGafxXiMDtMhCH9EJ5gy"}]}]}} {"custom_id": "Rangamati_12825_7695_13338_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PPBn4T8weLQ2iyvvKtHMmd"}]}]}} {"custom_id": "Sherpur_7182_3591_7695_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3ACrdB23NPFBy4UygCBKhh"}]}]}} {"custom_id": "Gaibandha_11286_11286_11799_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F1A8fxE5qXx3G7rDF2F9Hk"}]}]}} {"custom_id": "Magura_16416_14364_16929_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LPEA7ZzEYoaEMAUbYMbFC7"}]}]}} {"custom_id": "Panchagarh_18981_18981_19494_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EBiGEhg3emfXGaVsJsgFnz"}]}]}} {"custom_id": "Chandpur_19494_14877_20007_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-52gMSNCxwe4hpaoUUHpabq"}]}]}} {"custom_id": "Thakurgaon_9747_3078_10260_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7gj8mjw7fLYzmdohCVFwoo"}]}]}} {"custom_id": "Sunamganj_7695_11799_8208_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TbxowD32rgcEqrggddtdpk"}]}]}} {"custom_id": "Rangpur_25137_15390_25650_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RTVwXtH6wfTc4QniPdWfNj"}]}]}} {"custom_id": "Naogaon_29754_31293_30267_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5zVnqDTk2nNasUo1Jho6c2"}]}]}} {"custom_id": "Sylhet_11799_21546_12312_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M48zwXLVaoTQPb12Yy32xn"}]}]}} {"custom_id": "Tangail_14364_31293_14877_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QrGyJTraKENSWT3U7FaGov"}]}]}} {"custom_id": "Kishoreganj_21033_7182_21546_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TbLG8avr8w2oSHtBwWEmzE"}]}]}} {"custom_id": "Satkhira_15903_24111_16416_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NViTMeEJhuY565qgiVT11D"}]}]}} {"custom_id": "Bagerhat_11799_1026_12312_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ro7AeMTiNPiemUVdocS8iR"}]}]}} {"custom_id": "Rangamati_9234_26163_9747_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WKqyiNL2ctdQVALis3JpTt"}]}]}} {"custom_id": "Lakshmipur_9747_9747_10260_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D5JnmV8KwbwfXBVJVwhcU5"}]}]}} {"custom_id": "Mymensingh_10773_41040_11286_41553.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PJaEv9tP1onq7FAayBrCBK"}]}]}} {"custom_id": "Sherpur_5130_1539_5643_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WdYAWV7ompk7q8yrX67gAX"}]}]}} {"custom_id": "Jamalpur_4617_28728_5130_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AAWeofbaBmy4oJSPthwcGn"}]}]}} {"custom_id": "Dinajpur_40014_37449_40527_37962.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AB7yXtr7pXSREMt5raf585"}]}]}} {"custom_id": "Faridpur_12825_2052_13338_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QvPsqFHx7XbqfySws4mbw9"}]}]}} {"custom_id": "Mymensingh_22059_17955_22572_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4HM8CmVNPSuKLXAuAn1Qos"}]}]}} {"custom_id": "Comilla_18468_4617_18981_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JynUQ8zzWDsfKtt7z8L5jk"}]}]}} {"custom_id": "Jhenaidah_16929_18981_17442_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NWXm9TYp9Tz9qHRyWGwx1k"}]}]}} {"custom_id": "Rangpur_18468_30267_18981_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-G9jeMGiA5D1LqvnELBfJcK"}]}]}} {"custom_id": "Bhola_9747_45657_10260_46170.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7HgV6iCjHYMwnj9nrUxYHg"}]}]}} {"custom_id": "Kushtia_27702_22059_28215_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8qMvaLkidSL7NbtDQ9hwPK"}]}]}} {"custom_id": "Chittagong_34884_43092_35397_43605.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MYRCb3bs8PQSF7mVQ2wBe6"}]}]}} {"custom_id": "Barisal_13851_28215_14364_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7ETyDJTUQSc6vY2hZDLtYt"}]}]}} {"custom_id": "Maulvibazar_15390_22059_15903_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-B3YWYwvGrjbzo6S6AUW1Le"}]}]}} {"custom_id": "Khulna_7695_37449_8208_37962.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NW65QkggVDrkA2oAsNYJpg"}]}]}} {"custom_id": "Patuakhali_7182_11799_7695_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PGSRnKDneT3tyv8LcbyTNA"}]}]}} {"custom_id": "Netrakona_9747_25137_10260_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LmAr15s9VFcWp19B2i8zfQ"}]}]}} {"custom_id": "Bogra_13338_13338_13851_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5M41fmoDNqzn3UJ3ETQJ7M"}]}]}} {"custom_id": "Kurigram_3591_32832_4104_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9N99ELPSUWgkwBqsTnxP6W"}]}]}} {"custom_id": "Rajshahi_24111_23598_24624_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P6mdexD3h64sKfZNw9S3Sp"}]}]}} {"custom_id": "Khulna_21546_1539_22059_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8o3W1ir1p5ZjdxPD34TNx9"}]}]}} {"custom_id": "Chandpur_3591_1026_4104_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P49bGVvv6C8cq2DqU3zJ9H"}]}]}} {"custom_id": "Rajbari_24624_9747_25137_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VPF8JM6JB1RfBFSGUaEjVg"}]}]}} {"custom_id": "Lakshmipur_12312_26163_12825_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CT8K1YopfjYVfbg13PVt83"}]}]}} {"custom_id": "Thakurgaon_10260_14877_10773_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P9NzPKQBCaWNPCT8wAAp5N"}]}]}} {"custom_id": "Rangamati_18468_62586_18981_63099.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4DqE646gRbxXr5jMpBq9Wq"}]}]}} {"custom_id": "Nilphamari_5130_16929_5643_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2DTj5rxit3nyZGgSZn3qjW"}]}]}} {"custom_id": "Bogra_5130_12825_5643_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EzDTK6QRw6v34jf5go2bRa"}]}]}} {"custom_id": "Dinajpur_17442_15390_17955_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SCpetbfpqLo7dTF9uTsXih"}]}]}} {"custom_id": "Faridpur_10260_7695_10773_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Wjn1dWWtZNVyQRcBR87DKd"}]}]}} {"custom_id": "Khulna_9747_33858_10260_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3DVpXBWj79BJFwb4ac3vam"}]}]}} {"custom_id": "Noakhali_6156_2565_6669_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6FWUkNraUyAjhdV2wUgcpQ"}]}]}} {"custom_id": "Feni_4617_17442_5130_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WDmhCroJU8gCE1JBW8j1tJ"}]}]}} {"custom_id": "Tangail_4104_21546_4617_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XuQCycNLT8Mp84hTvWpXkA"}]}]}} {"custom_id": "Gopalganj_1539_7695_2052_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-57QzSwYx5h43fXp3LvASoi"}]}]}} {"custom_id": "Narsingdi_11799_17442_12312_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9PDRUgVsKdSn8nDv1Dy1vp"}]}]}} {"custom_id": "Faridpur_6669_8721_7182_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SdSVphBVQk8YB2AWbiau4V"}]}]}} {"custom_id": "Khagrachhari_11286_24624_11799_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NtzYBykQABwWQdYTZzML8N"}]}]}} {"custom_id": "Brahamanbaria_6669_20520_7182_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D1k3vawCBnPLqHgEXhc93T"}]}]}} {"custom_id": "Netrakona_25650_18981_26163_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Wnet5so2FQKRaMGH39DxCH"}]}]}} {"custom_id": "Mymensingh_8721_38475_9234_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7VHaRty1fhCXYtFV2YHzt9"}]}]}} {"custom_id": "Naogaon_22572_11799_23085_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QUo93Ey6U7MLSQdLSYAsA2"}]}]}} {"custom_id": "Bogra_15903_5643_16416_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JSYqz27DueUeJmeZB9nNry"}]}]}} {"custom_id": "Sylhet_9234_26163_9747_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2nEZgv3qu9SDf8Erch7YHj"}]}]}} {"custom_id": "CoxsBazar_7182_16929_7695_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JoPM9rG4J6NzHmu33tmziw"}]}]}} {"custom_id": "Bhola_6669_4104_7182_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ShAYejHhA3NcjbqQyaiYAQ"}]}]}} {"custom_id": "Rangamati_26676_69255_27189_69768.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ad7Mr9sshzUr6d1Pwn2cvF"}]}]}} {"custom_id": "Madaripur_12312_6669_12825_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VovsF7wVSg21jEwybXiujf"}]}]}} {"custom_id": "Gopalganj_12825_9234_13338_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Dh69SG7jbCW9AuTTp6HTc7"}]}]}} {"custom_id": "Lakshmipur_8208_5643_8721_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AhEZw8SiKRjRHQkNKfJhGo"}]}]}} {"custom_id": "Rangamati_13338_31293_13851_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5VD28NbprrCBtws57g7He2"}]}]}} {"custom_id": "Bandarban_23598_36423_24111_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1qN4kFjipMxsoyoZAdqWzE"}]}]}} {"custom_id": "Barisal_14364_10260_14877_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UyAnShW77Wm27k3hRP1hJy"}]}]}} {"custom_id": "Sirajganj_10260_24624_10773_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fqx1tNGdBG5fEi2wVzRyJH"}]}]}} {"custom_id": "Sunamganj_23598_9234_24111_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L3UEU7BCstty2emyA6YcE9"}]}]}} {"custom_id": "Rangamati_10773_50787_11286_51300.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-413bRFMXTjVFetrNNonkJV"}]}]}} {"custom_id": "Khulna_8721_49761_9234_50274.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AtNk9CEUKRxpYUjtMYDSNa"}]}]}} {"custom_id": "Jamalpur_10260_8208_10773_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LMCE2iGS4nXJ7wU41LBnpx"}]}]}} {"custom_id": "Sirajganj_26163_36423_26676_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GizWyzqD3svSgqciBWpy1o"}]}]}} {"custom_id": "Khagrachhari_16416_33345_16929_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PsfVEib26Ds2HWTwBcSWYL"}]}]}} {"custom_id": "Khagrachhari_7182_31806_7695_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QVVSa6vm5SXNQGjH6qv1Fq"}]}]}} {"custom_id": "Munshiganj_7695_3591_8208_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FZGGJ8ABBDq6ZZUWmJj5cb"}]}]}} {"custom_id": "Comilla_7182_10773_7695_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6tWVQSJAcDZn7ZJDJ1Uebc"}]}]}} {"custom_id": "Gopalganj_21033_14364_21546_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HbhkPXPk8faFcrD6kVyWxP"}]}]}} {"custom_id": "Magura_13851_14877_14364_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8to8dbRZobxjJ1vYR77y9k"}]}]}} {"custom_id": "Chittagong_21033_21546_21546_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3BPJxvAFTKyLFN8bG2WHAP"}]}]}} {"custom_id": "Jessore_26676_8208_27189_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8u2vUDW5YKxMndPVmKzB6F"}]}]}} {"custom_id": "Mymensingh_6669_33345_7182_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LgdjN55LUqw4WxY6BDb3Eg"}]}]}} {"custom_id": "Comilla_5643_14877_6156_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F215LpXnYD4UMTaeJMHuyS"}]}]}} {"custom_id": "Patuakhali_5130_17442_5643_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JnjJ4dnmBcpb2TzEHMeitT"}]}]}} {"custom_id": "Chittagong_6156_23085_6669_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5g27CBveqxCPWZEAB7EbAf"}]}]}} {"custom_id": "Comilla_19494_27702_20007_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DmzJi4Lx8zgCgjxHrQXUu2"}]}]}} {"custom_id": "Jessore_25650_6156_26163_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MjYZSGyXSuTNWNvV8FVZDK"}]}]}} {"custom_id": "Kurigram_19494_30267_20007_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5HzpxioKUvG9t7UjDp2fSZ"}]}]}} {"custom_id": "Rangamati_18981_33858_19494_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VMizouLUq277sJT6JDMkPV"}]}]}} {"custom_id": "Khagrachhari_17442_31293_17955_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GVRDobEQjKBXxpYesv3AJA"}]}]}} {"custom_id": "Mymensingh_17955_37449_18468_37962.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PyfK1uXbYd3g6vGyf4XXkc"}]}]}} {"custom_id": "Bandarban_11286_31806_11799_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PWduwUjtQ8SdvsmcziUAxD"}]}]}} {"custom_id": "Barisal_6669_5643_7182_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4L553kEXr4niBfvn7kPk9R"}]}]}} {"custom_id": "Sirajganj_21033_38988_21546_39422.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XzVua38LoYRaNgJTi6jsyW"}]}]}} {"custom_id": "Bogra_28215_5130_28728_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AnY3yrEadaieTcdaHfyWeK"}]}]}} {"custom_id": "Narail_14364_13851_14877_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WAejWXxEUwEhgjaGxJDkZ2"}]}]}} {"custom_id": "Naogaon_4104_7182_4617_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Xko4jqgLpLzJHL7wzJK6AM"}]}]}} {"custom_id": "Bagerhat_5643_17955_6156_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VoKX8LSvFF9CoTacdgc5p2"}]}]}} {"custom_id": "Khagrachhari_20007_30267_20520_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2kig8B2cxzu8AafZtdhXK8"}]}]}} {"custom_id": "Barisal_11799_13851_12312_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VCnn9oLGJf1TmTcSeuyz5S"}]}]}} {"custom_id": "Jessore_6156_8721_6669_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KzbDepvtzF3zdXUaetbXjn"}]}]}} {"custom_id": "Lalmonirhat_15390_24111_15903_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-58Nu3MacikyrxHGeGbC7ea"}]}]}} {"custom_id": "Chuadanga_5643_8208_6156_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2PNjUPujFo1VJLsJehEVLV"}]}]}} {"custom_id": "Chittagong_25650_13851_26163_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6G9eEWxHgqyTajQX8mLcok"}]}]}} {"custom_id": "Mymensingh_16416_9234_16929_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EMnZuv7cFFA8EdgV3wUrZm"}]}]}} {"custom_id": "Manikganj_20007_11799_20520_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9tLHV37KEMumDLsRDCGeVA"}]}]}} {"custom_id": "Pabna_17442_8721_17955_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MEamTyzWbQtFrBPrLCR9L9"}]}]}} {"custom_id": "Habiganj_12825_9234_13338_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EN2REAgWSpHCocoxFSxFwk"}]}]}} {"custom_id": "Natore_13851_25650_14364_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K9zLrihC9nf2xy7vcjUv9s"}]}]}} {"custom_id": "Bandarban_14877_7695_15390_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NSXCeYL9jTVNE8w25WiMkq"}]}]}} {"custom_id": "Barguna_6156_25137_6669_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EWNuaE96NvJPKwcEsCgVkP"}]}]}} {"custom_id": "Kishoreganj_26163_7695_26676_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1ahTS4c19BCAxyUMhKYwoy"}]}]}} {"custom_id": "Khulna_9234_60534_9747_61047.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2SynyhysReacXYzJfZLFQ5"}]}]}} {"custom_id": "Satkhira_13338_37449_13851_37962.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8i43Wr9oWbJnj2ydUQvfrs"}]}]}} {"custom_id": "Khagrachhari_13851_21033_14364_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SjXTvbSiQHUhaD34TGVbxj"}]}]}} {"custom_id": "Dinajpur_28215_26676_28728_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HcsA89SMNaEKvfuN8pXsyf"}]}]}} {"custom_id": "Pabna_20007_20007_20520_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LNR4wUyaSmF5tqwbVLBP5y"}]}]}} {"custom_id": "Satkhira_7695_44118_8208_44631.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T8Ph5Ce9yt1EM9o9y7k7qN"}]}]}} {"custom_id": "Chittagong_27702_22572_28215_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MCAksv3eESgnjrMRyqkJ9U"}]}]}} {"custom_id": "Rangpur_17955_26163_18468_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J22yGbokMassiQkG96i93B"}]}]}} {"custom_id": "Brahamanbaria_25650_10260_26163_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ESYHEpEyE1wfheB5R3gezE"}]}]}} {"custom_id": "Bagerhat_14877_20520_15390_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VVR2Sp6EeEw9FKYBcTEqEx"}]}]}} {"custom_id": "Narail_11799_6156_12312_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6bKbQTFQ5L7XJspc5pkCL1"}]}]}} {"custom_id": "Sunamganj_14877_21033_15390_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9b7v5kKGyCtRtqFjixbbpP"}]}]}} {"custom_id": "Nilphamari_9747_22572_10260_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XSq1NdDfcbdAHUR7jLjhSK"}]}]}} {"custom_id": "Bandarban_9747_38475_10260_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1W7e2B3JHnEYvT497JgdpD"}]}]}} {"custom_id": "Manikganj_15390_6669_15903_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XE1bQTUNWf5QWrZWtFrtWH"}]}]}} {"custom_id": "Kushtia_24624_16416_25137_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9YUBsYqf29fBnr5fp68PRi"}]}]}} {"custom_id": "Bhola_10260_16929_10773_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PCSERVhBfAdr5e8gN7U5Kb"}]}]}} {"custom_id": "Netrakona_21033_20007_21546_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V8TC5HvrFyntMzEgdcpU6J"}]}]}} {"custom_id": "Kurigram_2565_25650_3078_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3CKCmst2fJkhtyv6qGQWRM"}]}]}} {"custom_id": "Sunamganj_21546_4617_22059_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AxFjqmfrdYWt5TkmMQCxbm"}]}]}} {"custom_id": "Sunamganj_12825_15390_13338_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Awpa1YXSP3uU2BRURiexAn"}]}]}} {"custom_id": "Khulna_11286_39501_11799_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EiAdvDft44ATsAkQGYH4Xr"}]}]}} {"custom_id": "Bogra_21546_14364_22059_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VBNBLiUU8hoveQy2NgBt42"}]}]}} {"custom_id": "Sylhet_23598_9234_24111_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KNhEgCXC3eZXjfcG4FRfRj"}]}]}} {"custom_id": "Bandarban_17955_44631_18468_45144.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XpaHtiJkPXWsMfpiDgg8qc"}]}]}} {"custom_id": "Bandarban_18981_24111_19494_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JmRkMpaKEtZCrz7JUMAuxT"}]}]}} {"custom_id": "Khulna_11799_51813_12312_52326.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2QNDQjy9sgzCsu7YHBcwWc"}]}]}} {"custom_id": "Brahamanbaria_16929_3078_17442_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-78uSUgDpWLncChBkonbyWW"}]}]}} {"custom_id": "Bagerhat_15903_9234_16416_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F94Ngun82tjzpWnFrNoLjw"}]}]}} {"custom_id": "Satkhira_12312_51813_12825_52326.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NPzeNDw9234RBkojzriFD2"}]}]}} {"custom_id": "Chuadanga_4104_12312_4617_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GqM5irdG6TvjnKzyEKvMq1"}]}]}} {"custom_id": "Pabna_31806_25650_32319_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3RmkS1DqQjR737G5u6NtQJ"}]}]}} {"custom_id": "Comilla_23598_14877_24111_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VME8RtFENf1eK217vujT6a"}]}]}} {"custom_id": "Bagerhat_5130_55917_5643_56430.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MeBMCGgVu8czR3GD2XQU1i"}]}]}} {"custom_id": "Patuakhali_18468_4104_18981_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CJjb2CzK9p6neHbyA8GJZu"}]}]}} {"custom_id": "Narsingdi_8208_12312_8721_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W9ZtBbS3boSRRWzqN8Temh"}]}]}} {"custom_id": "Barisal_6669_9234_7182_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-17JdqgYYLNh3yKghAUuiPS"}]}]}} {"custom_id": "Sunamganj_25137_9747_25650_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Lo3eD8NuHdrCeNv8rMRpAM"}]}]}} {"custom_id": "Gazipur_5130_11799_5643_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9EF8Nr6kcEKoxq57YbPMSF"}]}]}} {"custom_id": "Rangamati_27702_86184_28215_86697.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9SdjKHDKcJjZ4rVJpsYnvu"}]}]}} {"custom_id": "Bandarban_27702_50787_28215_51300.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FTgYKJoZ8Xs8wToChkfz4K"}]}]}} {"custom_id": "Satkhira_17442_48222_17955_48735.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WBfeweUWSBnspLVhxi4RBu"}]}]}} {"custom_id": "Lakshmipur_8208_3078_8721_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RijQTeuynF8agv5wW4Z6Zj"}]}]}} {"custom_id": "Rajshahi_11799_15390_12312_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D4YZDzYvPme43M27J6ZjKe"}]}]}} {"custom_id": "Satkhira_14364_38475_14877_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RwT9abp5QFLGfpJHP7EcLg"}]}]}} {"custom_id": "Khulna_5643_22572_6156_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KUwTuB3PAQZR5GtZvGkQaB"}]}]}} {"custom_id": "Khulna_16416_15390_16929_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PDKHUTNASpWTwzGhRBrfbo"}]}]}} {"custom_id": "Rangamati_16929_52326_17442_52839.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SDxPBaCvzDey92QoqLhUXn"}]}]}} {"custom_id": "Mymensingh_13851_19494_14364_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7DkjWeZ3j59TfdJLju8U4q"}]}]}} {"custom_id": "Rangpur_11799_2052_12312_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RKstwXFRvHw8eALjhcYuW2"}]}]}} {"custom_id": "Sherpur_8721_20007_9234_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V4pC7mS1zoQzzexzf5x6ei"}]}]}} {"custom_id": "Joypurhat_9747_4104_10260_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NCBC8tC1ZskAQ8A2wXFBZd"}]}]}} {"custom_id": "Pabna_32832_15903_33345_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2smssvBchqrnWvfqg5n6XC"}]}]}} {"custom_id": "Lakshmipur_7182_12312_7695_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MvQb3dp3yzoGGVprjGaHA8"}]}]}} {"custom_id": "Gazipur_15390_8721_15903_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M62xxXppUtk8W6AHE3pipV"}]}]}} {"custom_id": "Jhenaidah_18981_9234_19494_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FJUykkZEPNYpSyjNzVhhjZ"}]}]}} {"custom_id": "Comilla_26676_18468_27189_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6K42XMqEK4n82U7gh8zsmX"}]}]}} {"custom_id": "Nawabganj_17955_12825_18468_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9ZF6Sjz458YKTmEmphUYTU"}]}]}} {"custom_id": "Sirajganj_13338_31806_13851_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JKMfJYJesY9QEPDUuXwpKy"}]}]}} {"custom_id": "Brahamanbaria_22059_3591_22572_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vqkij186XYEXxy2BvZ7Udj"}]}]}} {"custom_id": "Mymensingh_21033_4104_21546_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SEV74c2RxnLrSVmanx52AY"}]}]}} {"custom_id": "Comilla_7695_5130_8208_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L8M1z9yfQ5JgDW21ncxe8k"}]}]}} {"custom_id": "Khagrachhari_6669_28728_7182_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VMwE78qXrF2L3BjNAeE3jq"}]}]}} {"custom_id": "Khulna_4617_23085_5130_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MYANwKN6AWkwovWMKD7zmP"}]}]}} {"custom_id": "Bandarban_15903_25650_16416_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EgCGNByBAT4SHK6MLVzqEv"}]}]}} {"custom_id": "Pabna_29241_15903_29754_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5wxRz17HFzFDFvq6H6eZXv"}]}]}} {"custom_id": "Rangpur_18468_26676_18981_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QZT6dVmjsphE8k2xMTBnCu"}]}]}} {"custom_id": "Patuakhali_9747_15903_10260_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BiZeqtaWdPfLZi3M9BX96U"}]}]}} {"custom_id": "Jhalokati_2565_11799_3078_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K46vy1GmmA32baAphDWJBo"}]}]}} {"custom_id": "Manikganj_12312_9747_12825_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JK2wHouCCAMJUa1vE9Lkex"}]}]}} {"custom_id": "Rangpur_9234_513_9747_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AetLymWSbA438AJv3XcUrb"}]}]}} {"custom_id": "Patuakhali_5130_38475_5643_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2Z3BmmDUFFVTsuxHZnpF6Y"}]}]}} {"custom_id": "Bandarban_7695_21546_8208_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NsWmPyU8972ndL1bv76ccZ"}]}]}} {"custom_id": "Patuakhali_21033_17442_21546_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RqDUmrue8LysJLb6odY4s6"}]}]}} {"custom_id": "Bagerhat_3591_44631_4104_45144.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7HoxvWgneJG3tXjmgoTExJ"}]}]}} {"custom_id": "CoxsBazar_7695_12825_8208_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VSjRipocuuVBKMqd3Xo6Z7"}]}]}} {"custom_id": "Rangamati_27702_60021_28215_60534.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QMbXddSuU8pg3LmY2cmiCL"}]}]}} {"custom_id": "Bagerhat_4617_55404_5130_55917.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8Qq4Drx6XLKrz7n6NE2q4G"}]}]}} {"custom_id": "Comilla_28215_18981_28728_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9WuQRG5TkiCBQt55JpVuaV"}]}]}} {"custom_id": "Lalmonirhat_12825_23085_13338_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QspD27QeozEMnrDkXqhySU"}]}]}} {"custom_id": "Kurigram_18468_15390_18981_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BsfFy1cyKmugxC17TPnnny"}]}]}} {"custom_id": "Mymensingh_9234_10773_9747_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7bTLwCnk8rerezeH9hp7Fw"}]}]}} {"custom_id": "Rajshahi_14364_11286_14877_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VttiTcKeiWU5Ltxqqn9hvw"}]}]}} {"custom_id": "Dinajpur_12825_1539_13338_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NUf9W7vAX4TR2iJ3sz6uR3"}]}]}} {"custom_id": "Gaibandha_16929_11286_17442_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-82dgVjTauK88i6rTpcjpfs"}]}]}} {"custom_id": "Sirajganj_14877_29754_15390_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5opwv5QBeU1AuMBkSDzkH5"}]}]}} {"custom_id": "Barisal_13851_20520_14364_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SL5jSPRAL5dcy5vCkZAyi6"}]}]}} {"custom_id": "Pirojpur_9747_18981_10260_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JtHEPUb4XhReEyBXAkQnqd"}]}]}} {"custom_id": "Chittagong_16416_10260_16929_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Roo36ntca7Juj5Cfiqcv8m"}]}]}} {"custom_id": "Noakhali_5643_14877_6156_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GobpjEL8ndoCGqg1mRKCJe"}]}]}} {"custom_id": "Maulvibazar_22059_21546_22572_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HvsekuDnavKrKWReXxCtnv"}]}]}} {"custom_id": "Feni_11286_15390_11799_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A8iZ5mj7EqGCMsvD78HNYk"}]}]}} {"custom_id": "Bogra_10773_9234_11286_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NWfRFNZKedwo7Ea5LwGT56"}]}]}} {"custom_id": "Jhenaidah_20520_8208_21033_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HPJ8q1hZ2uK9zMh2dnhTP6"}]}]}} {"custom_id": "Gopalganj_12312_1539_12825_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JBsRVGJWDmxKRdj98Zp1Yq"}]}]}} {"custom_id": "Bandarban_12825_10773_13338_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6rXd5sNr6i6ohhBmB8A2YY"}]}]}} {"custom_id": "Jamalpur_8721_33345_9234_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BbZ1tzWt8aELJQoAzYBNR7"}]}]}} {"custom_id": "Thakurgaon_22572_7695_23085_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SDWE8LSTqFs7cTiEHQ1cra"}]}]}} {"custom_id": "Narayanganj_7182_1539_7695_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BF3gy8vAKvPG21T6rTMHH8"}]}]}} {"custom_id": "Kurigram_7182_20007_7695_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2p4dEKYywFLfoMDnfaBhMs"}]}]}} {"custom_id": "Narsingdi_513_17442_1026_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4nVM5WAtbfmrdBuy3ULrs5"}]}]}} {"custom_id": "Comilla_7695_17955_8208_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DYnVpcfsVpSuSLGfwMhoBT"}]}]}} {"custom_id": "Jhalokati_5130_15903_5643_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2myhXUnPhkhdprEv95k9ej"}]}]}} {"custom_id": "Rangamati_10773_0_11286_513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QSz4skJPrEaP5RBmYXniHd"}]}]}} {"custom_id": "Naogaon_5643_23598_6156_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-95CdCK6F9mBBrpAS4f78q4"}]}]}} {"custom_id": "Dinajpur_15390_27702_15903_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MBGDhVKDgZXQBkFXcfm1Ei"}]}]}} {"custom_id": "Sunamganj_33345_9234_33858_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8i6ctF6dR5ssGh7sQSm2QD"}]}]}} {"custom_id": "Bagerhat_3078_26163_3591_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-URyAz8aUdkU9pHfNC6Vvvy"}]}]}} {"custom_id": "Noakhali_8208_48735_8721_49248.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-52ipmu76bFMMQNXbTwfcwM"}]}]}} {"custom_id": "Shariatpur_2052_8721_2565_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FYDZm6Hn4Cpo6fhkAwmfdt"}]}]}} {"custom_id": "Khagrachhari_9234_16416_9747_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JGmZB3HjA2wF8jxKibuCWk"}]}]}} {"custom_id": "Rangamati_27189_77463_27702_77976.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T4c18ft1gBAiaNcsThQSMY"}]}]}} {"custom_id": "Nilphamari_9234_2052_9747_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7qhnDwxFoVBBaSWzWMiwTm"}]}]}} {"custom_id": "Tangail_3078_21546_3591_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LQgQRUNQ1z8MtmzeUt73dV"}]}]}} {"custom_id": "Rajbari_14877_6669_15390_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EeRfBcR9vDxXXpfxkHApJA"}]}]}} {"custom_id": "Bagerhat_5643_34884_6156_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BoJRzwR8RJJi6EHcSFDYx4"}]}]}} {"custom_id": "Jamalpur_10260_13851_10773_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9fVRwFvKv4HRhFXXSSQjJQ"}]}]}} {"custom_id": "Tangail_3078_25137_3591_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EBMRcHF91HH7fTT4Hdfgzo"}]}]}} {"custom_id": "Sirajganj_8208_27189_8721_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M3bxZ8UJMV9B3RsxMqrXvg"}]}]}} {"custom_id": "Bandarban_7182_35910_7695_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TUuuqU2DV35WY7Zkj1B8oQ"}]}]}} {"custom_id": "Pirojpur_6156_28728_6669_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Dfnfi79HrZKkV6Quebwzo6"}]}]}} {"custom_id": "Khagrachhari_12312_25650_12825_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XnwysPSBxwiregnwHCvABx"}]}]}} {"custom_id": "CoxsBazar_1026_17955_1539_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L7pPKV7rmQk5s3kjtbTyg5"}]}]}} {"custom_id": "Kishoreganj_2565_14877_3078_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JNhCwANhVM8bDSbHZHJ6VK"}]}]}} {"custom_id": "Tangail_12825_11286_13338_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WAtggBHJMFZdK3xURrPS36"}]}]}} {"custom_id": "Madaripur_14877_17442_15390_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MYdGrhiD5saYqeeckTZ5Da"}]}]}} {"custom_id": "Nilphamari_15903_7182_16416_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D8ZkvfJsMkHggUE4znAtHZ"}]}]}} {"custom_id": "Sirajganj_21546_38988_22059_39422.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LwMtnY9QK623EpMZkQ3EBN"}]}]}} {"custom_id": "Khagrachhari_5130_35910_5643_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JpooabFaf5jx4Esoi4TmD3"}]}]}} {"custom_id": "Gaibandha_20520_10773_21033_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SumxLk9aszADL29ge7FL5D"}]}]}} {"custom_id": "Mymensingh_22572_45144_23085_45657.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6VpmwideQXRB4JW67z53EW"}]}]}} {"custom_id": "Bogra_513_13338_1026_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JXkWkavC4JBq6PR1wTvzDh"}]}]}} {"custom_id": "Sylhet_10260_16929_10773_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TxgeRXNAXPbneVTZat8Arf"}]}]}} {"custom_id": "CoxsBazar_18981_25137_19494_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DugiQpXkPa1DGrstRxqE3i"}]}]}} {"custom_id": "Bogra_31293_19494_31806_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XE4eYq9KW6TTQxE5ajq1FJ"}]}]}} {"custom_id": "Mymensingh_21033_20007_21546_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BE5uoe7jTzevNG9kUUUb6E"}]}]}} {"custom_id": "Sylhet_7182_16929_7695_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3qdgnHzKeqguUHkkkFvAZ2"}]}]}} {"custom_id": "Dinajpur_10260_27702_10773_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UshiwSh5brPK4UTZeA2cqc"}]}]}} {"custom_id": "Gaibandha_24624_13338_25137_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1CiFFTY6Et11X414PEhWMs"}]}]}} {"custom_id": "Naogaon_21546_13338_22059_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8xJM37p3PYNTcoUcL34HwK"}]}]}} {"custom_id": "Khagrachhari_14877_43092_15390_43605.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DpFtD3GhhhXskut9RJPo7D"}]}]}} {"custom_id": "Sunamganj_23598_17955_24111_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gxhg3E4VhKMqBEznng324S"}]}]}} {"custom_id": "Sirajganj_9234_22572_9747_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A5XJFPBeBH9RkNWxFDfrtw"}]}]}} {"custom_id": "Khulna_1539_33858_2052_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XStcMd4yf3CcE8c5ZcGDLZ"}]}]}} {"custom_id": "Jhenaidah_15390_19494_15903_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VShYVXwN6PFnv3zWVzXs5k"}]}]}} {"custom_id": "Mymensingh_33345_32832_33858_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QQwSqq6WnDdVctyJnK3oYP"}]}]}} {"custom_id": "Kishoreganj_22572_13851_23085_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-71EiP9Xy6pB7cNei1NMrDg"}]}]}} {"custom_id": "Lalmonirhat_15903_24111_16416_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CSgCzjEeL5QZ6hu4UWwp68"}]}]}} {"custom_id": "Gopalganj_10773_5643_11286_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BcRfrpuYXqzJGh62WfGiuS"}]}]}} {"custom_id": "CoxsBazar_13851_24111_14364_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5sVCbN13pn12XQ5oPjoMQh"}]}]}} {"custom_id": "Gaibandha_12312_26676_12825_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BLzQB19Sp8sBSbjawhuhfW"}]}]}} {"custom_id": "Madaripur_9747_10260_10260_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N7WSUitufVXz7pJ1E32Zv7"}]}]}} {"custom_id": "Maulvibazar_15903_20007_16416_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NC6gHrZ3btjsgcqYAypYZf"}]}]}} {"custom_id": "Khulna_5643_61560_6156_62073.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S5wo2bsE6Rc2JtkgnvnVP3"}]}]}} {"custom_id": "Maulvibazar_8208_15903_8721_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TXqUDEpbVrr6Gs22wr6f1D"}]}]}} {"custom_id": "Habiganj_16416_1539_16929_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W5dhnAgkYrAgLuLgT5jfmc"}]}]}} {"custom_id": "Chandpur_6669_11286_7182_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8SQBz6dPhTHGvnjqAdi1Wt"}]}]}} {"custom_id": "Barisal_24111_15390_24624_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5ZyCPMFrK89DaZXLeSLpeX"}]}]}} {"custom_id": "CoxsBazar_2565_1539_3078_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1LLvX6hHKta1fXkpKKakWw"}]}]}} {"custom_id": "Chuadanga_9747_5130_10260_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8nivpUZmVza2BvSoAq9JDu"}]}]}} {"custom_id": "Bandarban_20520_15903_21033_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CKUw5R9L6zJ4DMi3n5cRUy"}]}]}} {"custom_id": "Bandarban_11286_28728_11799_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-B4NETbctx9bUkBJ4PwgmTQ"}]}]}} {"custom_id": "Noakhali_8721_22059_9234_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2yaCuYxfFSFaHe7rXQyBgJ"}]}]}} {"custom_id": "Sunamganj_9747_9747_10260_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XVuDELRAUht13grWt8DYto"}]}]}} {"custom_id": "Tangail_6669_35397_7182_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4KMe4U1Rre9KNspFhcKXto"}]}]}} {"custom_id": "Noakhali_9747_48222_10260_48735.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-E2VnbaCoCCTzgQvauGcVd3"}]}]}} {"custom_id": "Jhenaidah_12825_13338_13338_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SSYHiUo1uHbsAtwaDtyi9V"}]}]}} {"custom_id": "Sunamganj_21033_17442_21546_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HrQzSUwe3u6PThw8XtvzAp"}]}]}} {"custom_id": "Comilla_2052_6156_2565_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NzWJ5TQC4UNJhUEDe3Fwwh"}]}]}} {"custom_id": "Comilla_15903_20520_16416_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LJZPCbFWb4GbuVXRFveETc"}]}]}} {"custom_id": "Bagerhat_6669_61047_7182_61560.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VYHyeKXdgyZTxEtD8j2T8M"}]}]}} {"custom_id": "Jhenaidah_28728_3591_29241_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-364Fqv9Qn8gSQUmq5FQKvN"}]}]}} {"custom_id": "Rajshahi_513_11799_1026_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CoYKyQQWZ2XVA3ifEbcB6v"}]}]}} {"custom_id": "Rangpur_9747_513_10260_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SqHNc5AdpEzYyB3bmZafTB"}]}]}} {"custom_id": "Kurigram_16929_40014_17442_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EjT3CbJDQvquPWqyeVL1rH"}]}]}} {"custom_id": "Jamalpur_11286_17955_11799_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-249mA8kt5aM7EpCvZtThMT"}]}]}} {"custom_id": "Faridpur_15903_16416_16416_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1qaWoiQ1Vgt69MXAJZpT6L"}]}]}} {"custom_id": "Sylhet_26163_21033_26676_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EUV8WRwcHDd4UhAuXHyqjb"}]}]}} {"custom_id": "Kishoreganj_17442_16929_17955_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CifTuixddSDjFg9Mxtr38s"}]}]}} {"custom_id": "Sirajganj_22059_29754_22572_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CaGppskwzec2a7SDz84hPY"}]}]}} {"custom_id": "Brahamanbaria_15390_26163_15903_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VZM4ha3jNAg38pCuGZtrpv"}]}]}} {"custom_id": "Chandpur_13338_10773_13851_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UVjo9FPXB8PTFCCjN6rVYU"}]}]}} {"custom_id": "CoxsBazar_4617_18981_5130_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HN5zXR9w9afhD9vLHp1FLc"}]}]}} {"custom_id": "Shariatpur_3078_13338_3591_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PYsKZVA8LLnfPmPQVeFQfW"}]}]}} {"custom_id": "Mymensingh_24624_3591_25137_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N3zXXFmoWFTDxVztHXkaaB"}]}]}} {"custom_id": "Sunamganj_6669_9747_7182_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LWZ5zeuY77Rs93WKdVBwcX"}]}]}} {"custom_id": "Chandpur_17442_8721_17955_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WTVYh4vR5C335AgJZq5qi6"}]}]}} {"custom_id": "Sirajganj_14364_24111_14877_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Khwc1Cisdpm41oFcu3dobi"}]}]}} {"custom_id": "Panchagarh_5643_8208_6156_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NCs2jmDG7uJ7EoeCpLEhaF"}]}]}} {"custom_id": "Rajbari_1539_7182_2052_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FXqiYbLJ6oFXE1HA2b9EJq"}]}]}} {"custom_id": "Natore_2052_7695_2565_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7zCRDcGHroBxp8YBrw23oR"}]}]}} {"custom_id": "Sherpur_7182_2052_7695_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RQKuj5LdrpJyiMxiwxD1ZF"}]}]}} {"custom_id": "Munshiganj_8721_7695_9234_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3cAUiGF5jScdKAVbFAC6aN"}]}]}} {"custom_id": "Rangamati_25650_58995_26163_59508.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7BQ379ojCxUPr6kMr3sX4P"}]}]}} {"custom_id": "Munshiganj_16416_6669_16929_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SoWwoVrrvPf58UUtfAH3Dn"}]}]}} {"custom_id": "Dinajpur_12825_13851_13338_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2WRqPc5PxHqssZ5jhovA4r"}]}]}} {"custom_id": "Khulna_5130_27189_5643_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NopX98FR26TeSBjSktGVSe"}]}]}} {"custom_id": "Faridpur_26676_15390_27189_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9CYN2RAtPhHip2apL5Umcm"}]}]}} {"custom_id": "Rajbari_17442_13338_17955_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ENYWXdnsXwvW8Qix9edRoa"}]}]}} {"custom_id": "Rangpur_19494_30267_20007_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5LCkPEygbrwFZyhhAqraLU"}]}]}} {"custom_id": "Tangail_9234_36423_9747_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RDAVU744Z2mZqf8tLie8ET"}]}]}} {"custom_id": "Kishoreganj_9747_9747_10260_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1vX3TqVtP3iEb7hrqUGrr3"}]}]}} {"custom_id": "Bagerhat_7695_45144_8208_45657.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SkB9TGEs8naYsucqXxxtPp"}]}]}} {"custom_id": "Mymensingh_7182_22572_7695_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5KJkR2gJHt7hp8pJMrhqLd"}]}]}} {"custom_id": "Natore_18981_3591_19494_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4SqWmGJvymxSD6JfnZw5Gd"}]}]}} {"custom_id": "Shariatpur_5130_14364_5643_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9cVie6KWhZb8C2zvP59L8A"}]}]}} {"custom_id": "Rangamati_17955_17442_18468_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PBR3RZdGr7fZRb41gsg4Z3"}]}]}} {"custom_id": "Bandarban_7695_45144_8208_45657.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VCVbsCHZU8Zqi1sMaNNrj8"}]}]}} {"custom_id": "Tangail_21546_20007_22059_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TGqY9WJs8QXfZ1cijNwyys"}]}]}} {"custom_id": "Natore_15903_13851_16416_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UpxdsesLMaoP81M3EBezwm"}]}]}} {"custom_id": "Barisal_20520_22572_21033_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GDixHqKNievi2pLBBWjtE6"}]}]}} {"custom_id": "Maulvibazar_3078_28215_3591_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MUCfsech1p58P3MaVaVjFm"}]}]}} {"custom_id": "Bhola_17442_38475_17955_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YN67gAZijJFHAKTsYsN6ep"}]}]}} {"custom_id": "Chandpur_16416_11799_16929_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KXfve6WoQogQdJq3hM67yA"}]}]}} {"custom_id": "Nawabganj_9234_15903_9747_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7FES6J8W7MB22Ux7DVHdym"}]}]}} {"custom_id": "Jamalpur_10773_25650_11286_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VAv1PvGHpsJJWS6Xkb92iD"}]}]}} {"custom_id": "Tangail_14364_4617_14877_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RMrsWZ3o8LsY3xu9HYKcyJ"}]}]}} {"custom_id": "Satkhira_5130_17955_5643_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5kpxmc2e8ZiL3oc5yt9bgy"}]}]}} {"custom_id": "Kishoreganj_2565_13338_3078_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BiQHt4ybtioimLK6zG94g6"}]}]}} {"custom_id": "Kushtia_17442_22059_17955_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J5vuq7HTyiHMnuTgjWNCbU"}]}]}} {"custom_id": "Munshiganj_21546_3591_22059_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8Ly9Q8bKTBS1o1Eu9XdjwC"}]}]}} {"custom_id": "Chittagong_36936_38475_37449_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9nZRweb2RPEywei9EEzEJL"}]}]}} {"custom_id": "Noakhali_18468_18981_18981_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vf9QqazEVj5pHMBESLK8Dk"}]}]}} {"custom_id": "Sirajganj_18981_17442_19494_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UwS1zHEsTfzuTwLqRpX4u6"}]}]}} {"custom_id": "Narail_8208_4617_8721_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SWCdYgNMUs8rD39rdcdSFD"}]}]}} {"custom_id": "Madaripur_18981_20007_19494_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EBs4QLd2bWQsHR1RCPidfx"}]}]}} {"custom_id": "Shariatpur_1539_1539_2052_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-USHcbeiV4PH89pnkwcr7kv"}]}]}} {"custom_id": "Brahamanbaria_12825_11799_13338_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Cy8MpCcLvVZfgf3KiQnWEX"}]}]}} {"custom_id": "Lakshmipur_10260_21546_10773_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gff5gs4aCgnfWLurwSaY21"}]}]}} {"custom_id": "Bhola_13851_21033_14364_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DGjGmNQpDhzshwfc7kihKf"}]}]}} {"custom_id": "Habiganj_19494_27702_20007_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CiYm3UU8Sn3JPNY6kv49ih"}]}]}} {"custom_id": "Thakurgaon_15390_7695_15903_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PkXyh5CqiHZ8hbJSYUW9Ze"}]}]}} {"custom_id": "Narayanganj_9747_8721_10260_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ADXCPrS19CNWkVUyuCNsrZ"}]}]}} {"custom_id": "Bagerhat_8721_39501_9234_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2DwyFD2hpEDU8UtKPkMY5x"}]}]}} {"custom_id": "Thakurgaon_4617_8208_5130_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6pvCz3wvxDed1tvre3qwTc"}]}]}} {"custom_id": "Bogra_24111_9747_24624_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9DK4Jup1HfTFbV16jrZTAb"}]}]}} {"custom_id": "Bandarban_7182_3078_7695_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VrKVTt5ZkGHNtLLDqotKfx"}]}]}} {"custom_id": "Maulvibazar_6669_16416_7182_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1RqA64b5L6Arj3YtGPuMSS"}]}]}} {"custom_id": "Barisal_25137_8721_25650_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F6CJXdWqjVB5pTJe4AKuRy"}]}]}} {"custom_id": "Barguna_6669_9234_7182_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4n72verzsDFngCHADvpHPH"}]}]}} {"custom_id": "Rajshahi_19494_13851_20007_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HJqC4TiZ4kFgiU4yCxqzUo"}]}]}} {"custom_id": "Meherpur_513_15903_1026_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MEtuAfcF9hgmZyoFzhJbjR"}]}]}} {"custom_id": "Bogra_28728_15390_29241_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DL9xfzxr5J9NKU2RotreiY"}]}]}} {"custom_id": "Khulna_14364_16416_14877_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LMyzJMKJ4JUBbuexg6k4rj"}]}]}} {"custom_id": "Nilphamari_12825_15390_13338_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9K8auXd7gS7VVj6xqQkPs4"}]}]}} {"custom_id": "Netrakona_30780_21546_31293_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-G4h5ExHqGEF95WKf2BvAWr"}]}]}} {"custom_id": "Bandarban_3078_34371_3591_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XNk62aYfpLjaxsPUUhhsRg"}]}]}} {"custom_id": "Madaripur_13338_21033_13851_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QDovbbxZ1YCVh4yxP3wsBa"}]}]}} {"custom_id": "Rangamati_4104_61560_4617_62073.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3yEXH5bAB127YV7AsA3Xis"}]}]}} {"custom_id": "Feni_2565_23598_3078_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RAB4hvtcxmGEVr8sUWPZg6"}]}]}} {"custom_id": "Rangamati_18468_38988_18981_39501.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MoGAPF8nMP4zNHYmchdHSC"}]}]}} {"custom_id": "Madaripur_16929_18981_17442_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Lnh93o8z3MS3gT2dmo65dB"}]}]}} {"custom_id": "Sylhet_24624_16929_25137_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LHBH5gHDyQzcp1F9WNuF16"}]}]}} {"custom_id": "Joypurhat_4617_0_5130_513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LjTmZ2tM1SgtzvdWzSw3eM"}]}]}} {"custom_id": "Feni_7695_17442_8208_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UQTqNGjiKjcKyTwj4E63kc"}]}]}} {"custom_id": "Pabna_13851_20007_14364_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FxsYNQhe12npzxjKKJGDD6"}]}]}} {"custom_id": "Pirojpur_3078_21033_3591_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HYmRSmMBrNiD1FzoPg4fKd"}]}]}} {"custom_id": "Rajshahi_12312_6156_12825_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RJSz2sNQG8UtXbzmqjiysd"}]}]}} {"custom_id": "Maulvibazar_2052_15903_2565_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WD492L7VgLxQDeZE2ssd4G"}]}]}} {"custom_id": "Kushtia_23085_17955_23598_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7Ax4jEEckh2YH97jLVoQgr"}]}]}} {"custom_id": "Bandarban_20520_17955_21033_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5e6RKwxoHtt2DQBczBs6Ms"}]}]}} {"custom_id": "Bandarban_20007_23085_20520_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UWvCfJqQJ2f6fUKz71ccp7"}]}]}} {"custom_id": "Pabna_17955_20520_18468_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-96zL8MZrjWHpGE572r75Ax"}]}]}} {"custom_id": "Gaibandha_21033_14877_21546_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4TuXXXban3duGccu2tKJeR"}]}]}} {"custom_id": "CoxsBazar_9234_8208_9747_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UKay4KZD3Pv2FxMYBomDSk"}]}]}} {"custom_id": "Barisal_15903_8208_16416_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UiNpGgd7oPK8APm7zRAkgR"}]}]}} {"custom_id": "Jhenaidah_20007_23598_20520_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UMDgY338h81fuiyGcLUV4t"}]}]}} {"custom_id": "Thakurgaon_13851_5643_14364_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9GXwHUPgNTarcA3LCv8a8e"}]}]}} {"custom_id": "Bagerhat_2565_14877_3078_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X9ZuKWRMEptMqfrfLwp5cr"}]}]}} {"custom_id": "Maulvibazar_12825_8721_13338_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FBhnoZyUfv7UUH63Wug4rs"}]}]}} {"custom_id": "Munshiganj_10260_7182_10773_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LjEuptV1Na8HtEjNxGUN4Y"}]}]}} {"custom_id": "Kishoreganj_6156_7182_6669_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UK494Y6Zju1yDt3aFUZkcV"}]}]}} {"custom_id": "Barisal_4617_3591_5130_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U3Tg7kWmBUWSuf3Pq3M3R3"}]}]}} {"custom_id": "Kurigram_10773_16416_11286_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DSXnqRpoWecABUAHofPfAq"}]}]}} {"custom_id": "Shariatpur_11799_3078_12312_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3xunLZkcwtg16ccS84xqgU"}]}]}} {"custom_id": "Bandarban_7182_3591_7695_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5mCYx6Gr6bTbwzGXrm4eFa"}]}]}} {"custom_id": "Patuakhali_25650_6669_26163_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SYy4gvXGdgNtWxD1UFdNe2"}]}]}} {"custom_id": "Dinajpur_10260_14877_10773_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Q6JN2Wez1E88vMrXX9C4KH"}]}]}} {"custom_id": "Bogra_26163_7695_26676_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UBKJMWdcwA1RTSkSUyHjoS"}]}]}} {"custom_id": "Joypurhat_9747_5643_10260_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TN7qjApRgN1vDBBa1XZ26n"}]}]}} {"custom_id": "Mymensingh_27189_21033_27702_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MzNY446f8x8YXbkVQ246z5"}]}]}} {"custom_id": "Rajbari_513_9234_1026_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UWvaE5Fbchj1FWNiJVbTdZ"}]}]}} {"custom_id": "Netrakona_8721_18981_9234_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EvdMnZX6NzwjposYxdibS9"}]}]}} {"custom_id": "Bandarban_7182_11799_7695_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ar592cXGZnn9KqmHyZtjQZ"}]}]}} {"custom_id": "Kushtia_11799_11799_12312_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Q5wLqNF3Tx2QNXhkhPMhr1"}]}]}} {"custom_id": "Natore_21546_5643_22059_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5XvWC2tBBLHMADsPEqMxLm"}]}]}} {"custom_id": "Kushtia_20520_13338_21033_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CQTi4MkT39dhGJfJ6kqH3r"}]}]}} {"custom_id": "Jhenaidah_25137_12825_25650_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8LXezyxDWx3JSRy7dmcvMv"}]}]}} {"custom_id": "Panchagarh_6156_18468_6669_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-47YGL1SRogHHJFEy5kwios"}]}]}} {"custom_id": "Bogra_29241_16929_29754_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gi8CKAk4A6iupkh5mQum9p"}]}]}} {"custom_id": "Rajshahi_11286_15390_11799_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XgnbiayRnUL39QfuTH937i"}]}]}} {"custom_id": "Maulvibazar_1539_22572_2052_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WYQTNqZ2h2CvhvjEqECabA"}]}]}} {"custom_id": "Bogra_15390_9747_15903_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FtVXr8RYnjnmfQ1FyoHBLQ"}]}]}} {"custom_id": "Sirajganj_6669_18981_7182_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3A8Uo4oRsofsC6sMcs1SRa"}]}]}} {"custom_id": "Nilphamari_2565_18981_3078_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SBrSpRyJSgRUo9DvMDA67S"}]}]}} {"custom_id": "Narayanganj_2052_5643_2565_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VKK76qgkK562A1QRUAZVFS"}]}]}} {"custom_id": "Kurigram_9747_5130_10260_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N9ezK5nqsCjNuwq7r7GouJ"}]}]}} {"custom_id": "Jhenaidah_13338_17955_13851_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2Ci1LuWAfgSm2KRps4CWvz"}]}]}} {"custom_id": "Dinajpur_24111_35910_24624_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8jbVFzNPDEL9UzFUymmcf3"}]}]}} {"custom_id": "Lalmonirhat_7695_5130_8208_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XQfVqxZu5r1g8uEmxpk1S2"}]}]}} {"custom_id": "Sunamganj_27189_10260_27702_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VPLnQAMoxjawA9EAdHrgBE"}]}]}} {"custom_id": "Munshiganj_11286_12825_11799_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mgm8uFxHNuecojwFG86akj"}]}]}} {"custom_id": "Rajshahi_1539_9234_2052_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KfPhnSP3Ms4qFRmKmonU45"}]}]}} {"custom_id": "Patuakhali_8721_10773_9234_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EBma5Wi54qrNqeZVRedQz5"}]}]}} {"custom_id": "Jhenaidah_26676_4104_27189_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7hFkrpRuxF9TnmSf6RuaDv"}]}]}} {"custom_id": "Sunamganj_32319_21546_32832_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1Ew6oot1CF1KU1acgJZnp7"}]}]}} {"custom_id": "Mymensingh_9234_43605_9747_44118.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Wmft8DYHyNg4Q56fhgp5MS"}]}]}} {"custom_id": "Jhalokati_7182_3078_7695_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MieDvfv8wBiazQi5hG75cP"}]}]}} {"custom_id": "Bandarban_5130_38988_5643_39501.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Kq7sr5cE9JnuoyNZM9Zy1P"}]}]}} {"custom_id": "Jessore_9234_6156_9747_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3eNgvMwdmqEMr2WN8VuXzf"}]}]}} {"custom_id": "Satkhira_8208_6669_8721_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WbumNzDJXPjyuPcpnqKyAC"}]}]}} {"custom_id": "Bandarban_15903_31806_16416_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PNMsR8KxYoEH64P9VE3VvA"}]}]}} {"custom_id": "Bogra_7695_21033_8208_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SjBDLXoQaNdTLSJRaYkzZr"}]}]}} {"custom_id": "Mymensingh_21033_17955_21546_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NGDZYzE3tKXjd3HwcCo9R7"}]}]}} {"custom_id": "Bogra_26676_17955_27189_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KD2iJSz3K7tt2viKS4wBtM"}]}]}} {"custom_id": "Jhalokati_11799_5643_12312_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LPv7VuXD1ZFRmLALcQ1Zfh"}]}]}} {"custom_id": "Brahamanbaria_6669_19494_7182_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LuBbw1R4YB4wPVNTkVEGML"}]}]}} {"custom_id": "Sylhet_28215_17955_28728_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PEZgKc8QM7peCTHA5P7iCr"}]}]}} {"custom_id": "Kurigram_8721_18468_9234_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FGaHDN6L3kx6URSiPtj9ft"}]}]}} {"custom_id": "Maulvibazar_21546_12312_22059_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7DYCDiunXXbyYS8atiwP8S"}]}]}} {"custom_id": "Gopalganj_8208_5130_8721_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NVh4KYzfEsr3VzTRAs6Jo8"}]}]}} {"custom_id": "Tangail_3078_11799_3591_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9eKCy92pcZx82wt3guYd87"}]}]}} {"custom_id": "Rangpur_17955_17955_18468_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1Rt5GqpmA49HnyeR3ega8Y"}]}]}} {"custom_id": "Gaibandha_14877_26676_15390_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-B1CHkYAztKYnDaseat6EPx"}]}]}} {"custom_id": "Pabna_15903_21033_16416_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4HsnvaxvfYeV1g7nyn4UkJ"}]}]}} {"custom_id": "Faridpur_8208_22059_8721_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J5CVeBuuiN8fDTquDgkqRK"}]}]}} {"custom_id": "Thakurgaon_3078_11799_3591_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YGRcwzMU5bz3WHFfqbgTAQ"}]}]}} {"custom_id": "Joypurhat_12825_11799_13338_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5P4QcrVXQi4K8F4rPCDKEg"}]}]}} {"custom_id": "Netrakona_10260_3078_10773_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LfuvUJXDq84aAh3XdRNCf4"}]}]}} {"custom_id": "Rajbari_22572_6669_23085_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4sU3QA1ZhTLHq47HcQhZb7"}]}]}} {"custom_id": "Nilphamari_5130_20007_5643_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2ELnbofYueJAj1CRYqBJSn"}]}]}} {"custom_id": "Sunamganj_22059_16929_22572_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KkZ2JSo3MJ7sCEGwxn8YqV"}]}]}} {"custom_id": "Sunamganj_10260_11799_10773_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2FLauZn3FwjoNi3FaRNkHF"}]}]}} {"custom_id": "Mymensingh_30780_35397_31293_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WLL15CxgqYb9zWzoZJXxkM"}]}]}} {"custom_id": "Chuadanga_9747_4617_10260_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RzAqGg4xZv2wvHNRnJVhWQ"}]}]}} {"custom_id": "Chittagong_6156_19494_6669_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NsqVf1DB9HMBmstdfay88s"}]}]}} {"custom_id": "Gaibandha_6669_27189_7182_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NJ7bw7ZBntfJHX3dq8o9kB"}]}]}} {"custom_id": "Bhola_16416_16416_16929_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HnaD4PGaaU71G26aqoRfkF"}]}]}} {"custom_id": "Mymensingh_16929_5130_17442_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TUiX1xQMgcKeXAydBebxgH"}]}]}} {"custom_id": "Brahamanbaria_24111_9234_24624_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ErJnQaV7Pww6YcbuvxuwpN"}]}]}} {"custom_id": "Madaripur_14877_15903_15390_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7eoFyY4sgQByy62yCW48GH"}]}]}} {"custom_id": "Netrakona_10773_5130_11286_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SLBgy7Zd3aBbJGt5SPBiVR"}]}]}} {"custom_id": "Thakurgaon_10773_13851_11286_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SQDUdn6DjGK9eji4rsFkMU"}]}]}} {"custom_id": "Rangpur_13338_3591_13851_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X21k8fb8QhRaGX76W4i8f6"}]}]}} {"custom_id": "Bagerhat_5130_45144_5643_45657.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3JAp9mLEbZEjAjDYd4Ybjo"}]}]}} {"custom_id": "Rangpur_15903_33858_16416_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BbMhFMU2eu5iixV4d38neQ"}]}]}} {"custom_id": "Comilla_17442_2565_17955_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WV9QRsggcrMQSncMHf2ExM"}]}]}} {"custom_id": "Rangamati_9747_17955_10260_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2j4ANRHr33c8CZ4UPyV9R6"}]}]}} {"custom_id": "Barisal_4104_13338_4617_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6rPPdx6MxcPor6QYt3qcDh"}]}]}} {"custom_id": "Satkhira_6669_5130_7182_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JSi1kZMsikatQpUEB9rGwY"}]}]}} {"custom_id": "Mymensingh_23085_22572_23598_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3Y8RYjcPyFu5ohyengf7Mc"}]}]}} {"custom_id": "Chittagong_35397_42579_35910_43092.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Pu8tkxBJkLW5BGTWApRVjZ"}]}]}} {"custom_id": "Bandarban_7695_17955_8208_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HJw83SyzukY1eKGPveVPUH"}]}]}} {"custom_id": "CoxsBazar_7695_15390_8208_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PgxmQQBCwTbqPti1aVZe3R"}]}]}} {"custom_id": "Tangail_8721_26676_9234_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X57EViS1Fn7YtUzPdo7vSQ"}]}]}} {"custom_id": "Gaibandha_10260_24624_10773_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-REd2m6akpRJcNS3avrp3SZ"}]}]}} {"custom_id": "Sunamganj_25650_21033_26163_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vh8cA3qjxupCJoUBWsv5j4"}]}]}} {"custom_id": "Maulvibazar_13338_24111_13851_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-B846LVMWpJ1rSdZxRMAz9R"}]}]}} {"custom_id": "Sylhet_28215_13851_28728_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5cYmkJ8XJt17FSeyvJGVZf"}]}]}} {"custom_id": "Kishoreganj_18981_12825_19494_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5Sw3poDJ6Mbd1BLbq5QA2D"}]}]}} {"custom_id": "Nilphamari_4104_11286_4617_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Yc99q5YKeouoGdJmVj5Luk"}]}]}} {"custom_id": "Jhenaidah_24111_7695_24624_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fv4sAyHHWem8bQdi7JtQpQ"}]}]}} {"custom_id": "Rajshahi_14877_18468_15390_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LRPDecxAVSSwEiUz8Gj48W"}]}]}} {"custom_id": "CoxsBazar_12825_26676_13338_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-B51N6rNCZJgXmPfuQ3ar9n"}]}]}} {"custom_id": "Maulvibazar_24624_5130_25137_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LwupQ1mbGkGnCVa19SrhHr"}]}]}} {"custom_id": "Patuakhali_17955_21546_18468_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WGc9xsty88BwSRBajBUmLS"}]}]}} {"custom_id": "Satkhira_16416_6156_16929_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4iHBr8uoAtXxj7p17Amib6"}]}]}} {"custom_id": "Rangamati_15390_27189_15903_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QRLaTxR4ido39e2m7PT92N"}]}]}} {"custom_id": "Bandarban_25650_41040_26163_41553.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8bP21VgBk2tcUxAPvSRmRH"}]}]}} {"custom_id": "Sylhet_7695_27702_8208_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FVDQVCETrcNwva7AYc8FLT"}]}]}} {"custom_id": "Chuadanga_12312_4104_12825_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7E1W77FHbe3wxofDVdkQC7"}]}]}} {"custom_id": "Khagrachhari_5130_41040_5643_41553.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RmHM7YJyB4y3cqzjfXPRJP"}]}]}} {"custom_id": "Jhalokati_8208_10260_8721_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-25ySeT9m1UrS4rnjivBgJb"}]}]}} {"custom_id": "Chandpur_13851_15903_14364_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6QY2QAdGqAUgFV5GRm1xJy"}]}]}} {"custom_id": "Jessore_7695_6669_8208_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VPKEvr7NZqg1kkgzTnThgE"}]}]}} {"custom_id": "Feni_7182_5130_7695_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3Rek7bqwW9F8PRaoVyvVBw"}]}]}} {"custom_id": "Lakshmipur_4104_15903_4617_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Wuj8A5tx1xdidBuBPFD2AQ"}]}]}} {"custom_id": "Dinajpur_21546_28728_22059_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Dari4SL8eh4er96bSF1iXT"}]}]}} {"custom_id": "Kishoreganj_18981_12312_19494_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HSkdNPxSpV6aSkrAurm7uv"}]}]}} {"custom_id": "Netrakona_20520_19494_21033_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gt9EoHryhAEJhdC98nRfF4"}]}]}} {"custom_id": "Sherpur_7182_7182_7695_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EDxKpgdd6jC78dy41rTzC6"}]}]}} {"custom_id": "Narsingdi_8208_20520_8721_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JwzVhYFf3hkCKFJdZrv94s"}]}]}} {"custom_id": "Bogra_22572_4104_23085_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-88h8eJBzyRgPPa9jEtYghc"}]}]}} {"custom_id": "Rangamati_12312_6156_12825_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SG27EaaG5aeZgXBkVxxQh7"}]}]}} {"custom_id": "Satkhira_6669_10773_7182_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JQPdxQpfc8zK6uwiY18QLg"}]}]}} {"custom_id": "Madaripur_9747_12312_10260_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HaBG4foPyB6A5H7GfHg3Jt"}]}]}} {"custom_id": "Noakhali_8721_8208_9234_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7k16ZKMPsS4geeUxaVxnc6"}]}]}} {"custom_id": "Mymensingh_21546_45657_22059_46170.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EofAowGKQzQJBP3dxnyyPi"}]}]}} {"custom_id": "Jamalpur_14364_7182_14877_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YZwW6YrXSQ5ppQd1eEWzuj"}]}]}} {"custom_id": "Jessore_11286_10773_11799_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XuxNLDqr3uttxGpHj2GCFP"}]}]}} {"custom_id": "Sherpur_12312_16929_12825_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Y9M6UpgmBXC7wtH7Sookgx"}]}]}} {"custom_id": "Natore_8208_22572_8721_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SroExWW9S8H6pvEXNsjHyp"}]}]}} {"custom_id": "Gaibandha_22059_12825_22572_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X6hJWrj3HZrWXNKkQZPB3M"}]}]}} {"custom_id": "Nawabganj_6669_3078_7182_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EM6bp9kyR9xPPMJQRJCbg5"}]}]}} {"custom_id": "Chittagong_21546_27702_22059_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Qmqs9yQwAtQiKwdKm9aFMn"}]}]}} {"custom_id": "Maulvibazar_17442_24111_17955_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LD1kT5ckuZTAejRtK6VpZp"}]}]}} {"custom_id": "Rajshahi_24111_5130_24624_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R4zTiZmgP7SD27sBFUh5Fx"}]}]}} {"custom_id": "Feni_1539_11286_2052_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8b73qSediWtNEptUQN7zGo"}]}]}} {"custom_id": "Gaibandha_16416_12312_16929_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8Ami4sZ9TGXraYHY4koeCd"}]}]}} {"custom_id": "Barguna_9234_16929_9747_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LVRXrz7hX2K55uPtLiX27g"}]}]}} {"custom_id": "Sunamganj_34371_12825_34884_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gk5f98GyUy1qL1AV5kiogK"}]}]}} {"custom_id": "Kurigram_4617_16416_5130_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P3eGqXhWdjZHihUReRc3Bb"}]}]}} {"custom_id": "Sylhet_27189_11799_27702_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LsZm975sHxdVgSvyrckb5Q"}]}]}} {"custom_id": "Maulvibazar_8208_28215_8721_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7UCGHxXM5B9Ridc2Sv13UW"}]}]}} {"custom_id": "Chittagong_36423_26676_36936_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HS5tEdN6mfuYT7yVxDYXsU"}]}]}} {"custom_id": "Bagerhat_2565_59508_3078_60021.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2Vrd8DDJ21xCgjsHUJYBvf"}]}]}} {"custom_id": "Khulna_10773_25137_11286_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KvfzubFZkTqur86JJEHfPz"}]}]}} {"custom_id": "Magura_15390_12312_15903_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MofsMk3JVPF7GcoKRw7V4Z"}]}]}} {"custom_id": "Bagerhat_12825_48222_13338_48735.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KaAcXz1FP9995gmrbqvwDb"}]}]}} {"custom_id": "Brahamanbaria_27189_513_27702_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BELrEoijqFuimW7TwoZL1i"}]}]}} {"custom_id": "Madaripur_11286_9234_11799_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6m9tD65i3YHdZxhxB7s4Te"}]}]}} {"custom_id": "Bogra_16416_5130_16929_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Eh9mzGCxQLfW2z6tvtryBh"}]}]}} {"custom_id": "Nilphamari_4617_16929_5130_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CTeszxcwMTaVdYNaFuBXkU"}]}]}} {"custom_id": "Satkhira_8208_15903_8721_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-E8HBxfMVDbK4RzoMBS8wUA"}]}]}} {"custom_id": "Bandarban_15903_44118_16416_44631.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9GRymTNtfTc6aF6bYGsXad"}]}]}} {"custom_id": "Netrakona_12825_1026_13338_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QxmTmXThkVYNNZVvbz6g1h"}]}]}} {"custom_id": "Barisal_16929_12825_17442_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-165QudDoAwQmVWQBXEMHPx"}]}]}} {"custom_id": "Gaibandha_17442_16929_17955_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PN6zmTTFveQ2TDb684e6mm"}]}]}} {"custom_id": "Sirajganj_21033_3078_21546_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VadrsbGA8XDmM7wdkHQBp1"}]}]}} {"custom_id": "Chuadanga_9234_20007_9747_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Jm9vLd6ewEiZtpyzavfyDh"}]}]}} {"custom_id": "Netrakona_3591_17442_4104_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-C9kuT6bUdgtGwPu1KtrM5f"}]}]}} {"custom_id": "Dinajpur_17442_25650_17955_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Jbep6Q7Gsq2fQWNePRozyn"}]}]}} {"custom_id": "Pabna_27189_15390_27702_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PFE1ZaS3FcVqRqPeidBC4S"}]}]}} {"custom_id": "Brahamanbaria_7695_20520_8208_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DUDVgjWMx7V4kY2WnmmeQc"}]}]}} {"custom_id": "Comilla_25650_26163_26163_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8FFe9zPT2HqY3gBgPeQ5yL"}]}]}} {"custom_id": "Sunamganj_11286_14364_11799_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BqMESYeiWcFCVmWj2GFZCp"}]}]}} {"custom_id": "Kishoreganj_17442_21033_17955_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4tRr4xyh6cBGLC3PPrEh1X"}]}]}} {"custom_id": "Chittagong_23085_14877_23598_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BLpvTiF83VuBFLepZMdSVf"}]}]}} {"custom_id": "Mymensingh_11286_20007_11799_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GQQhwiEQvekvufSvQePndq"}]}]}} {"custom_id": "Rangamati_15390_66177_15903_66690.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Tm7JLRkLYKK89Kb7giChub"}]}]}} {"custom_id": "Naogaon_13338_2052_13851_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J2HfeZPiUe9MX1pNcr4hDX"}]}]}} {"custom_id": "Noakhali_16416_13338_16929_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3VxiUR6WFRLVJ3KRSboKJJ"}]}]}} {"custom_id": "Tangail_18981_16929_19494_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TEztj6qN19PWks1KsaYzs4"}]}]}} {"custom_id": "Panchagarh_14364_27189_14877_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LRpBuYtEygZQsCTYXADTbB"}]}]}} {"custom_id": "Barisal_3591_6669_4104_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LrE9NKqHVi26mw4Xi9CwAs"}]}]}} {"custom_id": "Jhenaidah_16416_16416_16929_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HCDHfnmHeH9GptnRnEeCWF"}]}]}} {"custom_id": "Rangamati_23598_65151_24111_65664.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R3ufHUgvGXc5NJLMmn9cZ1"}]}]}} {"custom_id": "Gaibandha_12825_10260_13338_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UZ3C8KcvC82ZtZQeTkA8Pi"}]}]}} {"custom_id": "Rangamati_9747_16929_10260_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J7mV1sirYYAJAKgQSgFB6S"}]}]}} {"custom_id": "Bandarban_3078_34884_3591_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HkYfgThLnUHmsu8Hk6CKES"}]}]}} {"custom_id": "Kurigram_15903_39501_16416_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SmidwG94LW3tYNCdXWKCki"}]}]}} {"custom_id": "Noakhali_10773_23085_11286_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W1u8XpfzKKFn3pZFchktdW"}]}]}} {"custom_id": "Brahamanbaria_26676_2052_27189_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-474as4BN3MQgFGhaAmQLxg"}]}]}} {"custom_id": "Madaripur_12312_15390_12825_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EsU5ic4NKrRo45zi8MyGvt"}]}]}} {"custom_id": "Mymensingh_10773_10773_11286_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2LCRCtd2Z4VRihVCHbsU5Y"}]}]}} {"custom_id": "Netrakona_16416_18468_16929_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1iC5S9qexLkKqr4LeAysUY"}]}]}} {"custom_id": "Gopalganj_7695_1026_8208_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AghuMW8EUniVBYaLj2A1Hr"}]}]}} {"custom_id": "Rangpur_12825_18468_13338_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XFPGqe2u5T96x6eqKn3qbo"}]}]}} {"custom_id": "Barisal_22572_25137_23085_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-54PT5VywCaRHxB32oUw2RL"}]}]}} {"custom_id": "Netrakona_16416_15903_16929_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FYLjFbXN1nao3rWj4karwJ"}]}]}} {"custom_id": "Patuakhali_5643_9234_6156_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6oWQBxGBnPHa6Q5nSUp7kx"}]}]}} {"custom_id": "Natore_16929_20520_17442_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ARHacyEfswL5GoKG5EKD2g"}]}]}} {"custom_id": "Mymensingh_8721_21546_9234_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ni4tQWRzCkaHVt1BsPYxjt"}]}]}} {"custom_id": "Nilphamari_5643_12312_6156_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KXKWJ2pzroCuTKSJzEQ1Cs"}]}]}} {"custom_id": "Comilla_28215_27189_28728_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RgnTD2BPDdCigksVUmzHLN"}]}]}} {"custom_id": "Shariatpur_11286_11799_11799_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GKh2vmDDVDLEux7HiYuJJF"}]}]}} {"custom_id": "Sirajganj_23085_36423_23598_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JKPdoxEAUcj9LqSn9BswTH"}]}]}} {"custom_id": "Chuadanga_5130_17955_5643_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SvjoJbD2oVUqB7mbXFTi5p"}]}]}} {"custom_id": "Chittagong_9234_29754_9747_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4Ht6e1tqtFrtC1tWVwN4sG"}]}]}} {"custom_id": "Rangpur_12825_4617_13338_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-H2aCC6cGsgSMQAqDXPPBza"}]}]}} {"custom_id": "Chittagong_34884_51300_35397_51813.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QTs6pgxstePgLSyuwssa58"}]}]}} {"custom_id": "Chandpur_6669_12312_7182_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3MJwLBr1ennkiy2bhTZSYC"}]}]}} {"custom_id": "Bandarban_13851_37962_14364_38475.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9rWfa1mJXEq8oKuskkMxbe"}]}]}} {"custom_id": "Shariatpur_2565_5130_3078_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LZ9SQWzwCqwTuX7k45jaW4"}]}]}} {"custom_id": "Kishoreganj_28728_9234_29241_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Tb75z1zn2Vg4VyGTg1qKtT"}]}]}} {"custom_id": "Chandpur_12312_4617_12825_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6f4qnbzRHetCeJJovape84"}]}]}} {"custom_id": "Narayanganj_11799_9747_12312_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WxcsGSkcQuCeeHjW64DtYv"}]}]}} {"custom_id": "Natore_8208_6669_8721_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EmbDpyWcB1h4CkpyzBCfSc"}]}]}} {"custom_id": "Rangamati_16416_52839_16929_53352.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WrzRMw2HWpSUPbnrCS2WqM"}]}]}} {"custom_id": "Khulna_4617_34884_5130_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ako46XgfqL17ecnJtnvYd2"}]}]}} {"custom_id": "Sirajganj_17442_32832_17955_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1bVAo3DVXfczKHF2zPnjjU"}]}]}} {"custom_id": "Nilphamari_12825_22572_13338_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Q9EW6GJcAtnGZkSCNEWibY"}]}]}} {"custom_id": "Rangamati_11286_20520_11799_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-B31pmbQJkmsKRq9e2V7Dpz"}]}]}} {"custom_id": "Tangail_12825_29241_13338_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2TqcCpAgCWQQmwnb9z3jTM"}]}]}} {"custom_id": "Jamalpur_11799_26163_12312_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BKTSPjsriSVPdF6K52Y8NK"}]}]}} {"custom_id": "Narayanganj_7182_0_7695_513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DQr9Xrfr7qcAEq8DB6m5V7"}]}]}} {"custom_id": "Naogaon_29241_11286_29754_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KDNuXRSiL7YTK8rpcdQQTM"}]}]}} {"custom_id": "Manikganj_7695_4104_8208_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PnkoeUyk5sLvGkk67SzKQq"}]}]}} {"custom_id": "Narsingdi_9234_1539_9747_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EK9J32zmGatfe55wTYx2wg"}]}]}} {"custom_id": "Sherpur_10773_5643_11286_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F8Kqi6A3vzdypVWaPrpo8V"}]}]}} {"custom_id": "Sylhet_35910_12312_36423_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mp4dVyWzBMTkoLva5JHoEt"}]}]}} {"custom_id": "CoxsBazar_21033_50274_21546_50787.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FtzFhUyPYAfmTVSJCjaeYs"}]}]}} {"custom_id": "Sirajganj_23598_24111_24111_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vtq1c1nR3dggawfrnNtSW2"}]}]}} {"custom_id": "Lakshmipur_12312_17955_12825_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-12gHibtE1KyWp3yjtqLryf"}]}]}} {"custom_id": "Comilla_26676_19494_27189_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Cej9r2L1GmrWYmGt7kBK92"}]}]}} {"custom_id": "Kurigram_9234_30780_9747_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VruhNFmzWfA3mhGg7iRHAf"}]}]}} {"custom_id": "Satkhira_12825_29241_13338_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-77NMNNCP8hDUtxbWu6wTdT"}]}]}} {"custom_id": "Natore_16929_8721_17442_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KPFPeUZ6BjhkxRpLTkTffU"}]}]}} {"custom_id": "Comilla_4617_15903_5130_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8StcmGmPBes29Vajfq2JJi"}]}]}} {"custom_id": "Noakhali_3078_16416_3591_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JusRhEJXby5FnT8rpKRH4N"}]}]}} {"custom_id": "Thakurgaon_10260_6669_10773_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BFQtQak6eD4bTZA25WGwZ9"}]}]}} {"custom_id": "Habiganj_15903_2052_16416_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bf1u3f1znYAgDpBcNjNt7G"}]}]}} {"custom_id": "Bandarban_16929_15390_17442_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TGQhbD8eoVGzMmY36uT4vR"}]}]}} {"custom_id": "Chandpur_15903_8721_16416_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MapYk2L4o4ckQK7kZAkp7s"}]}]}} {"custom_id": "Nilphamari_2052_13338_2565_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3WaPrA9pvHhA8SfBEGP6DQ"}]}]}} {"custom_id": "Rangamati_20007_20007_20520_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1LNK5725TGTXie5ZiQjvMX"}]}]}} {"custom_id": "Rangamati_9234_24111_9747_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QX6kohbBNggxh9mbboniS3"}]}]}} {"custom_id": "Nawabganj_7182_2565_7695_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X4wncBa6JC5oNwcUaKJtfp"}]}]}} {"custom_id": "Satkhira_6669_35910_7182_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CC37V9XMVh48ASZtDt4J4w"}]}]}} {"custom_id": "Panchagarh_12825_24111_13338_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JtpABQygcV5jdTk8fehkT3"}]}]}} {"custom_id": "Nawabganj_15390_19494_15903_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3rEyQipmpwdmEuw9wULchT"}]}]}} {"custom_id": "Comilla_25650_29754_26163_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XiV8rrf1pKzG2WUHJGfYnG"}]}]}} {"custom_id": "Kurigram_17955_12312_18468_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-664vtoxhifYDRC8bXLimCi"}]}]}} {"custom_id": "Chittagong_36423_30267_36936_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9Sg2wjRR3zYf8DFu2VgRh2"}]}]}} {"custom_id": "Kurigram_9747_2565_10260_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RRBxsVLYmNXaruMsABB8Le"}]}]}} {"custom_id": "Nawabganj_9747_4617_10260_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XwAAy3ekCPMHL7HBfPj9yz"}]}]}} {"custom_id": "Mymensingh_11286_13851_11799_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6kWiq6w4gLCC14fB39MWiQ"}]}]}} {"custom_id": "Lakshmipur_8721_21546_9234_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GrakoFtP4YQxTgGnVdfctz"}]}]}} {"custom_id": "Netrakona_18468_17955_18981_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WBW6VySzZ1q4ZLiPFdNCSr"}]}]}} {"custom_id": "Satkhira_4104_23598_4617_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-C4txzfUCp9V8L149rfzM5T"}]}]}} {"custom_id": "Chittagong_32832_32832_33345_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3S2agsaB4bqpVcchz3oC3t"}]}]}} {"custom_id": "Bhola_2565_12312_3078_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5XVSmPB73WVeUaSpi2aQhp"}]}]}} {"custom_id": "Chittagong_34371_48222_34884_48735.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9D3JQ47ZK3HSKvpMCCPQKG"}]}]}} {"custom_id": "Kurigram_17442_36936_17955_37449.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SyvaQ312hpf4bL7Ht6Hkev"}]}]}} {"custom_id": "Sylhet_21033_12312_21546_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AHXTXfpxka1HLiHCp5fMku"}]}]}} {"custom_id": "Gopalganj_12312_13338_12825_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ACWzwBuVTDWJn13wFVJUBo"}]}]}} {"custom_id": "Patuakhali_19494_26676_20007_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M63nCNFxMbAtknBPFj7RNB"}]}]}} {"custom_id": "CoxsBazar_7182_25137_7695_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WoFiimLPAwfFGNbCBFcc2F"}]}]}} {"custom_id": "Habiganj_17955_12825_18468_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EUbwK9BEpRVRmpffRqRYF1"}]}]}} {"custom_id": "Kurigram_5130_21546_5643_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LprNTDnskHCGFe5rWnPDLg"}]}]}} {"custom_id": "Maulvibazar_5130_33345_5643_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2RWX7hTfQ9hmWwbKDwnCfM"}]}]}} {"custom_id": "Bogra_22572_22059_23085_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K2Uf27wPjssPLWEqmKZy6D"}]}]}} {"custom_id": "Faridpur_18981_2565_19494_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7cwFREhfGB9kw8zUeZ2WLj"}]}]}} {"custom_id": "Patuakhali_2565_30780_3078_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Cj8L9GiXdG41FJvX6tTZ6L"}]}]}} {"custom_id": "Rangamati_6156_52326_6669_52839.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4kjLtHXgfGEq2wBuom86sP"}]}]}} {"custom_id": "Nawabganj_10773_22059_11286_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6rtirsV6mUe3dAfP2wgH3m"}]}]}} {"custom_id": "Maulvibazar_513_12312_1026_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1PcL4sVVCqJSBVTpU4EwPK"}]}]}} {"custom_id": "Shariatpur_1026_4617_1539_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Y7foFXWdLR2wDeQrHcam7f"}]}]}} {"custom_id": "Shariatpur_12312_13338_12825_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PaTFSBeTL5iLX16ChuXsgW"}]}]}} {"custom_id": "Bogra_19494_4104_20007_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6B5EpFGvazJsYD6vLyPwDe"}]}]}} {"custom_id": "Mymensingh_16416_22059_16929_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EK52QN2eQrhBcQHx37yLGB"}]}]}} {"custom_id": "Chittagong_19494_17955_20007_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CNj4qazdGwwvsj9jhyg1TL"}]}]}} {"custom_id": "Joypurhat_9234_11286_9747_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AC1SgDvHsyj7roVk6mgXKX"}]}]}} {"custom_id": "Khulna_3591_26163_4104_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CQmLVfPSHG6AhAT5NC6XVa"}]}]}} {"custom_id": "Narayanganj_7695_11286_8208_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LQjGhD8VgFuWeJdZf4WZ79"}]}]}} {"custom_id": "Barisal_22572_17442_23085_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HKu31KkH335f2mkGcbj7HC"}]}]}} {"custom_id": "Shariatpur_5643_6669_6156_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KfTbaQWFr6kihoGmCVvsPN"}]}]}} {"custom_id": "Joypurhat_14364_10773_14877_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VSX6rGipGMWbckSXMK88zT"}]}]}} {"custom_id": "Naogaon_13338_22572_13851_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GUzBkGPexq9BtP63Cxv31x"}]}]}} {"custom_id": "Sylhet_3078_6156_3591_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PpqBC5Dnj2wC4PmaDquyKw"}]}]}} {"custom_id": "Kurigram_2565_14877_3078_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LgAXSoxuCvXmZjd6N88Gzy"}]}]}} {"custom_id": "Khagrachhari_14364_15390_14877_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5zdqdcjNPsiiuwx5uu7JPw"}]}]}} {"custom_id": "Jessore_20007_21546_20520_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CykdTbENWUF5Xe7rUN98md"}]}]}} {"custom_id": "Bagerhat_1026_59508_1539_60021.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2k4L6oPyBniWd7uYbfW8xD"}]}]}} {"custom_id": "Bhola_15390_16416_15903_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NDVaw7AyapJgTf2NRj7Pfs"}]}]}} {"custom_id": "Maulvibazar_9234_26163_9747_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UfAggroH6ovNvY1rdzHPT9"}]}]}} {"custom_id": "Thakurgaon_7182_17442_7695_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AQeBYWoLhyrMCiE7ia5Vbd"}]}]}} {"custom_id": "Jessore_6669_9234_7182_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VW9HSFhxrEgZwChfirfSNu"}]}]}} {"custom_id": "Bogra_31806_21033_32319_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ADMsgGmPgrT4qkVqq1Wno7"}]}]}} {"custom_id": "Maulvibazar_3078_21033_3591_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5yuYZCxvnHxAMiY5HibQYh"}]}]}} {"custom_id": "Sylhet_9234_25137_9747_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JjDk5hiiY26atRaXLfwiuE"}]}]}} {"custom_id": "Pabna_32832_16416_33345_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NqRgnWw5QQXFGyHRYM4okL"}]}]}} {"custom_id": "Kishoreganj_13338_18981_13851_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P7kL1xviNmxpgZqYKTWr1m"}]}]}} {"custom_id": "Dinajpur_6669_7182_7182_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3JcPVsw6TW5pBz1EAwXMFQ"}]}]}} {"custom_id": "Shariatpur_5643_14364_6156_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R7h9ayYLv5xbeHTf8Bxjyh"}]}]}} {"custom_id": "Khulna_8208_12825_8721_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FEf5RFZG2JJ1RjTdGC4bYe"}]}]}} {"custom_id": "Sherpur_12825_5643_13338_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KE8qxyX7DemFhyLeGDjzDc"}]}]}} {"custom_id": "Bhola_12312_36423_12825_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1ZNQMkQB577JTgYidMBgoe"}]}]}} {"custom_id": "Naogaon_26163_31806_26676_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hgy11WEKRyckUHzXCJwYJx"}]}]}} {"custom_id": "Bandarban_19494_33345_20007_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Eeg96CPGFT98auQNSpaTtJ"}]}]}} {"custom_id": "Chittagong_21033_16929_21546_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NQziCXSEpgPWyxcr3z2A6g"}]}]}} {"custom_id": "Dinajpur_38988_37962_39501_38475.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CCmqwranSdGQKBnUm6jB54"}]}]}} {"custom_id": "Khulna_1539_29241_2052_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ThPtwGMfCp1x82jPdUMfiX"}]}]}} {"custom_id": "Kushtia_7695_6669_8208_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1ch82aKDLfuxShynW6r3w8"}]}]}} {"custom_id": "Naogaon_16416_20007_16929_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GgJT8RopVYaBvjxhm2zGsD"}]}]}} {"custom_id": "Kurigram_3078_25650_3591_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KgZ9Y6P12wH3UYNRtmaT6k"}]}]}} {"custom_id": "Sylhet_30267_14364_30780_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JK6iprGXLUNYFgx8Cv1eut"}]}]}} {"custom_id": "Gopalganj_10773_16416_11286_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Qi4SQoEdUzoswzUvmPQhns"}]}]}} {"custom_id": "Lalmonirhat_24111_18468_24624_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LejdTFdJy7nuVt2TDSfaCr"}]}]}} {"custom_id": "Chittagong_9234_6669_9747_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PgBANKUJBpHADgXfh1Prac"}]}]}} {"custom_id": "Gazipur_5130_7695_5643_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fc7w9cH9KH1tq4kUPKRRWh"}]}]}} {"custom_id": "Gopalganj_12312_3078_12825_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LvutmUq83qnK7CdJRDyeH4"}]}]}} {"custom_id": "Kurigram_10773_3591_11286_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DvM1EhKrLoUZNdVkKcQqmi"}]}]}} {"custom_id": "Netrakona_34371_27189_34884_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Nuc1foufJreSnKptjxYRn2"}]}]}} {"custom_id": "Gopalganj_513_11799_1026_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7Qpxzg6LmPeeXfJm5DGRuF"}]}]}} {"custom_id": "Habiganj_8208_18468_8721_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BK2jzVMa4QhETLkQvfBHdZ"}]}]}} {"custom_id": "Bhola_11286_35397_11799_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-35VpCfrArov1krVADTQSmf"}]}]}} {"custom_id": "Nilphamari_1026_2052_1539_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CBasqDHPH5MTbAs3fk53pv"}]}]}} {"custom_id": "Brahamanbaria_4617_22059_5130_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4sCTnaB41FCfjnZD6tT6wr"}]}]}} {"custom_id": "Tangail_21546_30267_22059_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7BoGwibvAsGwGtFQAuTTue"}]}]}} {"custom_id": "Barguna_11286_7695_11799_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-13ajEGVEeUJSkyJUBL9hEG"}]}]}} {"custom_id": "Mymensingh_26163_29241_26676_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WWzYqxzddMocaZwkK4qJpN"}]}]}} {"custom_id": "Sylhet_22059_4104_22572_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AkjoHCwvtbuMmCrQTUwnra"}]}]}} {"custom_id": "Rangamati_12312_55404_12825_55917.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2NqBYY3seHmteLeRyMUTTB"}]}]}} {"custom_id": "Noakhali_4617_9234_5130_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SP5ETfigXGw4GET58cTPgG"}]}]}} {"custom_id": "Manikganj_24624_12825_25137_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-78wRmd9ozFKKBKSChsUASP"}]}]}} {"custom_id": "Rangamati_18981_27189_19494_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6kiV3EoK4MMtxSr3NQ9Hu8"}]}]}} {"custom_id": "Sunamganj_5130_7695_5643_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NbJYpoVnqGFTxjhgUYz6ds"}]}]}} {"custom_id": "Khulna_0_22059_513_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5mGbFmUR58VWuGegSTQUXV"}]}]}} {"custom_id": "Chittagong_36936_27189_37449_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MR2C59Vcrg3Db5395TWykq"}]}]}} {"custom_id": "Patuakhali_10260_10773_10773_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LGQYBU6mCQQ6Ydpg3w3xkd"}]}]}} {"custom_id": "Bagerhat_10260_1539_10773_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Sr3g5RMwALBRxbP7V7FrRt"}]}]}} {"custom_id": "Faridpur_18981_18981_19494_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-E3ZYpYDPxXscUHEpCMKCZB"}]}]}} {"custom_id": "Sunamganj_15390_9234_15903_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EwizvUuZHYEBSbEN1qnBDj"}]}]}} {"custom_id": "Tangail_9234_36936_9747_37449.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8pptsXmsYEwXn8VzSJEGcG"}]}]}} {"custom_id": "Bagerhat_8721_28215_9234_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4YitoRBBkgg9z6GzBN55Fy"}]}]}} {"custom_id": "Chittagong_11286_25650_11799_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VyDofw4813kXa5RagCsEfy"}]}]}} {"custom_id": "Rangamati_11286_11286_11799_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PFc2G8B7iFT1kgLt2vHHnn"}]}]}} {"custom_id": "Joypurhat_2565_9747_3078_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MPEDxczmcGu9HZ1Eoemt8g"}]}]}} {"custom_id": "Sunamganj_27702_20520_28215_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PghjiQfx7UQy68M6E56r2P"}]}]}} {"custom_id": "Khulna_12825_34371_13338_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AtrPwPFmCfaYkDNmCuLwMi"}]}]}} {"custom_id": "Jhenaidah_23085_2052_23598_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5o8ybWYRvvi3SXfuBJXXpP"}]}]}} {"custom_id": "Comilla_21033_28728_21546_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-B1UA6yApywsjXBzFhp45fj"}]}]}} {"custom_id": "Panchagarh_10773_14364_11286_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Dwc3s3JBx5MfR1aUbVmLyu"}]}]}} {"custom_id": "Khagrachhari_7182_25137_7695_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2ULqvdXzs83vQKMrwcQDCN"}]}]}} {"custom_id": "Bhola_2052_33858_2565_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-H6R5RsDfeSHB5ZFmWuSYDs"}]}]}} {"custom_id": "Naogaon_5643_14877_6156_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SUoYHMiAj5hmaE4ANidxFe"}]}]}} {"custom_id": "Lakshmipur_12312_3591_12825_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9KyrkgF9daMDMntRou5E1h"}]}]}} {"custom_id": "Habiganj_19494_16929_20007_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BaGec8kTLETo1HiFZNi96q"}]}]}} {"custom_id": "Rangamati_23085_51813_23598_52326.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P8g6jRZx4tG38egRDUTtfz"}]}]}} {"custom_id": "Comilla_25137_32832_25650_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TdATMk2zcFKukmtEgL3Hn3"}]}]}} {"custom_id": "Pabna_18468_8721_18981_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JGxtxPsDqCHh8biwJWZC7a"}]}]}} {"custom_id": "Rangamati_12312_10260_12825_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7mPRLcaUZV7b6XnkQK2bAt"}]}]}} {"custom_id": "Chittagong_27189_46683_27702_47196.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ANvS2Gaq3wGAxThgJJtRqy"}]}]}} {"custom_id": "Netrakona_9747_26676_10260_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5kuC6BkjPZwfhNb6Hek9e5"}]}]}} {"custom_id": "Comilla_6156_7182_6669_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CW7oG7cNpFtHBaxTfpzd3r"}]}]}} {"custom_id": "Natore_7695_15390_8208_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ByJZ8XvsW6w1EZMQx7yQCF"}]}]}} {"custom_id": "Pabna_18468_16416_18981_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ei6QXgZwGG6227N1H2Ai83"}]}]}} {"custom_id": "Patuakhali_8721_6156_9234_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WZ4Q7uuC77rrAGLQYo6xw5"}]}]}} {"custom_id": "Kushtia_24111_12312_24624_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XJipLPLyowGavDWuPPD6nE"}]}]}} {"custom_id": "Sirajganj_17442_31806_17955_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TLkJ8eiwETur69qTuLBt7W"}]}]}} {"custom_id": "Magura_9747_8721_10260_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LiZzD3FQ4gYLqJLT73xybV"}]}]}} {"custom_id": "Thakurgaon_10260_4104_10773_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VPAV7djKXmhjqG7M8HBe62"}]}]}} {"custom_id": "Tangail_6669_8721_7182_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CAdwExNsDWiZTbNfuEmaXK"}]}]}} {"custom_id": "Habiganj_7695_22059_8208_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T5X6wN7ib3MCJCVNSwpm5i"}]}]}} {"custom_id": "Netrakona_13851_6669_14364_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2N5MuqEWP6fqkFZU8zDzEY"}]}]}} {"custom_id": "Jessore_7182_12825_7695_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S5YV2GvfJqbGPX57VDgnsd"}]}]}} {"custom_id": "Lalmonirhat_19494_25137_20007_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1BLPxaQYDHS8Cfr7RT8LpF"}]}]}} {"custom_id": "Magura_13851_11286_14364_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9zEkXmcXQjk8WhbyFJFpuo"}]}]}} {"custom_id": "Gazipur_10260_13851_10773_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4zWeJxaemsVNNR98j48roS"}]}]}} {"custom_id": "Bandarban_3078_35910_3591_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9WVHThXaFvGXNEdmKm5bJn"}]}]}} {"custom_id": "Narail_11286_14877_11799_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5LjPdAiGunj7Hoz3cD4zT1"}]}]}} {"custom_id": "Chuadanga_8208_7695_8721_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GxpCWHYScwCx1Mhav8uNb4"}]}]}} {"custom_id": "Pabna_4617_14364_5130_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fv3Q7j8RQ55FyAbewMw3Pi"}]}]}} {"custom_id": "Jhenaidah_20520_3078_21033_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7SJtMhYN4yXn4S8G91UFaz"}]}]}} {"custom_id": "Kurigram_9747_4617_10260_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9Cvzvm69EGTFeWuXXgFnVJ"}]}]}} {"custom_id": "Magura_15390_14364_15903_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RfundJggBcNeBUuDuqqp5H"}]}]}} {"custom_id": "Bogra_5643_15390_6156_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1EtYKtd65RNMwTMHaWQaNK"}]}]}} {"custom_id": "Mymensingh_22059_3078_22572_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WeX87QGCWuvXXU9cMkZqPS"}]}]}} {"custom_id": "Faridpur_513_7695_1026_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AWFoVRrvg5UUXPVf6s98vT"}]}]}} {"custom_id": "Thakurgaon_7695_16929_8208_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HCWZd8k9CKGdjAn2QgxT1N"}]}]}} {"custom_id": "Mymensingh_23085_37962_23598_38475.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JXUMkCQg7v6weZdtL5UzjN"}]}]}} {"custom_id": "Chittagong_29754_55404_30267_55917.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X2MCUp18sNrCBAiSGfnE5y"}]}]}} {"custom_id": "Rajbari_3591_2052_4104_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3giVnVCf4J4Z5Ky3VfohXZ"}]}]}} {"custom_id": "Satkhira_10773_44118_11286_44631.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6idSSyZgzusH7zZxVzkkjH"}]}]}} {"custom_id": "Patuakhali_11799_32319_12312_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A1VGcwA8jEs94bP5UjdVkx"}]}]}} {"custom_id": "Dinajpur_32832_28215_33345_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FMoEswNZ85CEVgm2FJN2iV"}]}]}} {"custom_id": "Magura_5130_2052_5643_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Xodpe9z9SePrjLAZkQbkN5"}]}]}} {"custom_id": "Khagrachhari_2565_46170_3078_46683.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DCn5u7dgx7gZLEx8TrRmc6"}]}]}} {"custom_id": "Mymensingh_7695_30267_8208_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U2UvFUoXEoRvs77CY7ExN1"}]}]}} {"custom_id": "Comilla_9747_15903_10260_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PwwYp1mm9t26B9CbfKYxjr"}]}]}} {"custom_id": "Khulna_8721_60534_9234_61047.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NHvRqgeZWUAffKYxVn4Yyf"}]}]}} {"custom_id": "Naogaon_32319_28728_32832_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MoUrag1iUBwY694tCqtGcs"}]}]}} {"custom_id": "Mymensingh_15903_39501_16416_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AK3fXZFboBBKX3F9Uvcq12"}]}]}} {"custom_id": "Sunamganj_27189_8208_27702_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NLnNt28tcwUyVKxcP38nDQ"}]}]}} {"custom_id": "Narail_11799_14877_12312_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8sxvpp8Wv2jAMiFozmgMmD"}]}]}} {"custom_id": "Sunamganj_11799_11799_12312_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D1CtFNRndhM9Qe88XXJkv6"}]}]}} {"custom_id": "Dinajpur_24111_27189_24624_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1cD9x52jiVSbgmHYkRuhSK"}]}]}} {"custom_id": "Barisal_22059_11799_22572_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FXKM7Xx4D4AXZ6hDev6oxZ"}]}]}} {"custom_id": "Khagrachhari_4617_41553_5130_42066.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EwQfBPZYLWva3TjPmR365N"}]}]}} {"custom_id": "Rangamati_30267_71307_30780_71820.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QkVb9XGRFkxNRfUffUWRFF"}]}]}} {"custom_id": "Patuakhali_16416_14364_16929_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JLmdB7hAEh3yxTeRVB5RCT"}]}]}} {"custom_id": "Bandarban_13338_17955_13851_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PpmywQ7muT7L2Emqw87B2m"}]}]}} {"custom_id": "Bandarban_8721_35910_9234_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RAnfLksB1Su7Sn9cGceNWK"}]}]}} {"custom_id": "Naogaon_30267_24624_30780_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K53efCQNR6vPm3qYSC2tz5"}]}]}} {"custom_id": "Bhola_3078_37449_3591_37962.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JzSz15CFrZKRY3GS6juBHR"}]}]}} {"custom_id": "Tangail_4617_28215_5130_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-86qf96wwYvBZhwJQDJh2Gy"}]}]}} {"custom_id": "Rajbari_17955_16416_18468_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Q97zj82SUpHkg1Y5xEfJEx"}]}]}} {"custom_id": "Magura_11799_11799_12312_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6pZdkB8yWYrEJ6KeRgZxR9"}]}]}} {"custom_id": "Sylhet_30780_5643_31293_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MdqfFGM76rn6L9rc8t7v9Y"}]}]}} {"custom_id": "Jamalpur_5130_32319_5643_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EA8PVs7TeHjBVS4UKLBiLC"}]}]}} {"custom_id": "Jamalpur_8721_21033_9234_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7a14vr1oKe2up8jBti7VxG"}]}]}} {"custom_id": "Gazipur_8721_14877_9234_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Kpmy6bwRtAGeVFxqgTcFj7"}]}]}} {"custom_id": "Thakurgaon_8208_5130_8721_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GmTRP3Fg96Uwg9ZXf26vV2"}]}]}} {"custom_id": "Lakshmipur_11286_14364_11799_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SsaAbgFDECuC1wGTs4y8Wk"}]}]}} {"custom_id": "Sunamganj_11799_8208_12312_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FBt1jwXQArtFFn4ciSiJC4"}]}]}} {"custom_id": "Patuakhali_3078_15390_3591_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GhTyy7YUKkctoaBsXofeyy"}]}]}} {"custom_id": "Khulna_11286_11286_11799_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-63SYThFZbVC2smWTkKzoCX"}]}]}} {"custom_id": "Patuakhali_20520_8208_21033_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BYYcMydsZD1EWPnGGjPycp"}]}]}} {"custom_id": "Kurigram_13851_14364_14364_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HPGkriX5wsrCZRqhAsKgfd"}]}]}} {"custom_id": "Khulna_7695_5130_8208_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JpbRuv9wqv4ESF2pgTnj1Y"}]}]}} {"custom_id": "Narayanganj_4104_6669_4617_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JyemnfRy8HYqPm8Nh4BJXD"}]}]}} {"custom_id": "Naogaon_7695_5643_8208_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QHtqEuvHXMJTBR5AmK49jc"}]}]}} {"custom_id": "Bogra_23085_2565_23598_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6hZoBF9EMUMbWfghLnqiab"}]}]}} {"custom_id": "Jhenaidah_21033_2565_21546_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S5nq8wZqcXD2PtW76u2UXk"}]}]}} {"custom_id": "Netrakona_15903_25137_16416_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RxEHeXzCPs4pw3YnxywWqH"}]}]}} {"custom_id": "Jhenaidah_15903_2052_16416_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EUZUs4twMwB6haDDBLEp5B"}]}]}} {"custom_id": "Rajshahi_27189_14877_27702_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6Wkc7zwhopoeBSkyr2VeLK"}]}]}} {"custom_id": "Netrakona_15390_29241_15903_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-51JyF82bDbdYKC76dePoEz"}]}]}} {"custom_id": "Dinajpur_22572_35910_23085_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5FLT13ACrP8QAjX9J8b4aK"}]}]}} {"custom_id": "Jamalpur_6156_17442_6669_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6bn35iA396TmqGJ7Zs2ko9"}]}]}} {"custom_id": "Dinajpur_27702_33345_28215_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LMUakvXBu5dM2eRdx2SPdJ"}]}]}} {"custom_id": "Bogra_27189_25137_27702_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-69L5kGDTn9KG3uK6snSRvS"}]}]}} {"custom_id": "Natore_22572_7182_23063_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PnC6713PX9dKKq679r5mDb"}]}]}} {"custom_id": "Pabna_30780_14364_31293_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FzRVvGFwDs71EW164ic8Y9"}]}]}} {"custom_id": "Chittagong_20520_14877_21033_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J2YDPW5aSWwxcR436LBJ8i"}]}]}} {"custom_id": "Pabna_16929_13338_17442_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VhBVRgqdgFkYKpVfywz5Li"}]}]}} {"custom_id": "Nawabganj_16416_16929_16929_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VKbQZLLAjM54tUT5S2TmRu"}]}]}} {"custom_id": "Kurigram_4617_30780_5130_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2RXux1XC84o6SEuhrnZC64"}]}]}} {"custom_id": "Faridpur_4104_8721_4617_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QGpRx8tLr3Wjj7tqrpF8Zi"}]}]}} {"custom_id": "Sirajganj_17442_18981_17955_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8xtTD7goz14Aewoh9SiCqW"}]}]}} {"custom_id": "Munshiganj_16416_9747_16929_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KfX5Xhd7zBps4G5CaaXZAX"}]}]}} {"custom_id": "Bhola_8208_35910_8721_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-How46GLhGQsQQ2KH3bvBx7"}]}]}} {"custom_id": "Chittagong_30267_23598_30780_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2juPRbxkpHSmM7cXyamkg1"}]}]}} {"custom_id": "Thakurgaon_10260_15390_10773_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VN2zYuy4DdrjGg57Ki7Nxv"}]}]}} {"custom_id": "Rangpur_17442_27189_17955_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BG5U3jPZAktdMBxefJz5vE"}]}]}} {"custom_id": "Netrakona_21033_21033_21546_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PZxa9v6X7dy3XRQ3R2Kc9E"}]}]}} {"custom_id": "Thakurgaon_13851_25137_14364_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VMqdiqaAN9AM3eBBADVGnV"}]}]}} {"custom_id": "Kurigram_4104_30780_4617_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FYhYW82GcoXTfuHzuE8gwv"}]}]}} {"custom_id": "Maulvibazar_12312_12825_12825_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ATcxdmp29yeFEpkAC59Ktx"}]}]}} {"custom_id": "Natore_11286_17442_11799_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6zP1fv15AThsdKa9ifqgfL"}]}]}} {"custom_id": "Gaibandha_10773_28728_11286_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2WuRK83cd3ie7wPKWVzgyB"}]}]}} {"custom_id": "Narail_13338_8208_13851_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9MYT76PzsWdZVckP26PWXj"}]}]}} {"custom_id": "Patuakhali_15903_9747_16416_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Q2gjiooLJL4G3MAnytFasq"}]}]}} {"custom_id": "Bandarban_14364_32832_14877_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QjhbK4bktHADrAiaL7DzTe"}]}]}} {"custom_id": "Narsingdi_7695_7182_8208_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J3kzGtYgE8u5DsmEVBxBhg"}]}]}} {"custom_id": "Satkhira_9747_43092_10260_43605.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XNQtFSMrzxubfJbjMUCv5v"}]}]}} {"custom_id": "Tangail_1026_17442_1539_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2jGavuwvEw8VcrgsMeMBWY"}]}]}} {"custom_id": "Khulna_5130_25137_5643_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-95XEeJwQi2BatJNMBcwFDr"}]}]}} {"custom_id": "Rajshahi_20007_3078_20520_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2zvqnihyE4PNqzb7kPQyKZ"}]}]}} {"custom_id": "Naogaon_2565_5130_3078_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4CnYQY64EygmtHYaYxLTs5"}]}]}} {"custom_id": "Brahamanbaria_17442_1026_17955_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MPRCkdkQi2oKQwyHvkFmZr"}]}]}} {"custom_id": "Sunamganj_16929_6156_17442_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ux3oCANdcJGT1MHPnfJdgU"}]}]}} {"custom_id": "Sirajganj_11799_24624_12312_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ca7H5rwqx6gsBJxLFX7rz2"}]}]}} {"custom_id": "Netrakona_20007_3591_20520_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mxg7J5bcB3SQ59ovAYuJdu"}]}]}} {"custom_id": "Rajshahi_29241_4617_29754_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EoUzMQXS8C2o7hxEwpg4nr"}]}]}} {"custom_id": "Kushtia_14877_10773_15390_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X1CVD7WhYjdy88xwSPaX2Q"}]}]}} {"custom_id": "Rangpur_15390_4617_15903_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GBag739uQnvYPLFqnqt4nm"}]}]}} {"custom_id": "Comilla_25650_25650_26163_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P1MwdBBvKtPk3NXGXfa5P5"}]}]}} {"custom_id": "Joypurhat_513_4104_1026_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KfoBoZbZQkMZGmro8EPBg8"}]}]}} {"custom_id": "Mymensingh_15390_12312_15903_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R5cdkgrqL1YRwvoCVJg6Yw"}]}]}} {"custom_id": "Tangail_11799_29754_12312_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-62MFdpsTH1uxyXP8cUa7Gq"}]}]}} {"custom_id": "Pabna_18468_9747_18981_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BWbktuTqy7BjpKdK8VfXMz"}]}]}} {"custom_id": "Gaibandha_2565_22572_3078_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7N5v8afCDgxMjvWddwzzgK"}]}]}} {"custom_id": "Lakshmipur_13338_12825_13851_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MpTf3zJBozWBUq7swAWT6h"}]}]}} {"custom_id": "Noakhali_17442_18468_17955_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-F1u36XRnwoHExbU3xGNCvP"}]}]}} {"custom_id": "Sunamganj_33858_3591_34371_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BgwwKnGgYPH2o8Ts5s44iZ"}]}]}} {"custom_id": "Bagerhat_13338_31806_13851_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MBQW5vi4ShcAoJErNC3P48"}]}]}} {"custom_id": "Jessore_16416_7695_16929_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RznAYceWwz3S4WaQrxR1Do"}]}]}} {"custom_id": "Lakshmipur_10773_11799_11286_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XvpaJ34rttMU74DCaMeBRy"}]}]}} {"custom_id": "Nilphamari_9747_15903_10260_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9WvuQwB9cQUQQhwHJqiYbc"}]}]}} {"custom_id": "Bhola_18981_35910_19494_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Li8zgvkmiSvysMwnGfhG5P"}]}]}} {"custom_id": "Habiganj_17442_30780_17955_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LUfjuLiTDB4njjsY9Ro9dC"}]}]}} {"custom_id": "Rajbari_9234_14877_9747_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CQE1fqTpzyWBh2L46rLQej"}]}]}} {"custom_id": "CoxsBazar_2052_16929_2565_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TFwSUWoHR17dyAebHU6kxS"}]}]}} {"custom_id": "Sunamganj_34371_10260_34884_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XewyMW3Vox4whL8FwMBKji"}]}]}} {"custom_id": "Nilphamari_11286_7182_11799_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BH2TvEvCdfif7MP9ZPdCGJ"}]}]}} {"custom_id": "Khulna_15903_24111_16416_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9zgAgqpQXwSavRSQoM8VWQ"}]}]}} {"custom_id": "Chittagong_18981_4617_19494_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EhJeSpvowx3Ba9DwSayYPD"}]}]}} {"custom_id": "Khulna_8721_26163_9234_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K48TsydinjVDKuCd2NzHjY"}]}]}} {"custom_id": "Chandpur_9747_20520_10260_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8XZvp5Pz4mvVPYQGcdPgFP"}]}]}} {"custom_id": "Noakhali_3078_31806_3591_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XyP8D1chNH2TxQnEDUy7vC"}]}]}} {"custom_id": "Chittagong_36423_35397_36936_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1x1JxkuYeGsw8mTXGmGpPs"}]}]}} {"custom_id": "Bandarban_16929_44631_17442_45144.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TCYAxGCB1BHtSVRdogQekn"}]}]}} {"custom_id": "Chandpur_13851_4617_14364_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Nm7P5LNyELgGjAnVtToLNV"}]}]}} {"custom_id": "Jessore_2052_15903_2565_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GSKm6dgiZbRistMQoeMgSt"}]}]}} {"custom_id": "Tangail_17955_28215_18468_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V2uCN82pmmRkKmiZEUgKft"}]}]}} {"custom_id": "Jamalpur_5643_12312_6156_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Tdrkies9CR7gSXCLrM6Zuu"}]}]}} {"custom_id": "Dinajpur_25650_35910_26163_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T4cXAYFY72V2JM1ZzmTEbT"}]}]}} {"custom_id": "Barisal_12312_28215_12825_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PJTYp8u7xwfhTFvpTLWn6o"}]}]}} {"custom_id": "CoxsBazar_13338_41553_13851_42066.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EV3CfEyk4tZvQgxyxfGz3j"}]}]}} {"custom_id": "Rangamati_17442_67203_17955_67716.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YUJ3u6kWiyj3d5PpZT5sDM"}]}]}} {"custom_id": "Jessore_15390_15903_15903_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RqpCgXRLZbKrpFFdFgfnA6"}]}]}} {"custom_id": "Tangail_13851_27702_14364_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8v8jJkUy4Mv6TtVPGPYTzQ"}]}]}} {"custom_id": "Rajbari_3591_5130_4104_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7d5sr7Lgov12iFwSYrXkHG"}]}]}} {"custom_id": "Lalmonirhat_10260_18468_10773_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SKjUvsiRVBD1rZqQMa84Sn"}]}]}} {"custom_id": "Panchagarh_2565_4617_3078_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-13DZjWzYNeTjExFVS2mk1d"}]}]}} {"custom_id": "Kurigram_11286_29241_11799_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UD7dyKnULPFx2NWusBtBde"}]}]}} {"custom_id": "Sherpur_12825_17442_13338_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AXnGgMRsvcKQh6p5KcjyNC"}]}]}} {"custom_id": "Natore_15390_15903_15903_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5pNvcNsgnXdgkmfQrvsAtK"}]}]}} {"custom_id": "Bogra_6156_15390_6669_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J4pMovQhvPE6Gya238eZct"}]}]}} {"custom_id": "Khagrachhari_18981_28215_19494_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KKVR6cyjYbU11FFwT8DiYW"}]}]}} {"custom_id": "Jessore_3591_12312_4104_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MFpK3GssYevnPFS2pLEvT4"}]}]}} {"custom_id": "Noakhali_9747_45657_10260_46170.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4njifqnabM86xjUnsAjpc8"}]}]}} {"custom_id": "Habiganj_18981_15390_19494_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5PVDdyGtR88PXN8jnBmhqy"}]}]}} {"custom_id": "Comilla_13851_18468_14364_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NcGXH33tLTVcRJPhjJzraS"}]}]}} {"custom_id": "Tangail_6669_40014_7182_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VBLLB487WfDDtqsMdD3AHu"}]}]}} {"custom_id": "Satkhira_9234_41553_9747_42066.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3M2mHRgabruzjGu5pAtWiN"}]}]}} {"custom_id": "Rangamati_10260_50787_10773_51300.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MgYAmru1crUnC4hncC4Buv"}]}]}} {"custom_id": "Kurigram_7695_25650_8208_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9k8cvxxJzHFMFVvU29fjGe"}]}]}} {"custom_id": "Narayanganj_10260_11286_10773_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MYtPpxqpF98VcbC1JpYhyT"}]}]}} {"custom_id": "Mymensingh_7695_38475_8208_38988.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fecxh5AnrC8BqsCa1Dm1hF"}]}]}} {"custom_id": "Bandarban_12312_9234_12825_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HypKapTf26CGd993eeZJX4"}]}]}} {"custom_id": "Bagerhat_6156_12312_6669_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-16sv1eDdTfQmEpyP1hMyJS"}]}]}} {"custom_id": "Shariatpur_12312_10773_12825_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VzKitukY8YJe3g9YnJTcba"}]}]}} {"custom_id": "Comilla_21033_32319_21546_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HJ6mFKXkekwtUCNWY7y8zg"}]}]}} {"custom_id": "Sunamganj_26163_19494_26676_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3hdHdBK754qegqEEsVUozu"}]}]}} {"custom_id": "Sunamganj_16929_14877_17442_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BpCNvy2qkKfiBJr52p62H2"}]}]}} {"custom_id": "Comilla_16929_8208_17442_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4nJ9KTv2FQwkL2TNa1EKMC"}]}]}} {"custom_id": "Bhola_4104_32319_4617_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RLG8D4xJosRhbcy32wwcf4"}]}]}} {"custom_id": "Comilla_23598_27189_24111_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WXQz4oQzwtCyYK6cRuneLf"}]}]}} {"custom_id": "Jamalpur_12312_6156_12825_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JwpEd4z5Cit6v1cD7q9Y2m"}]}]}} {"custom_id": "Bandarban_9747_6156_10260_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DihDKwnpLA9vfyWakbV83z"}]}]}} {"custom_id": "CoxsBazar_5643_14364_6156_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JKH3Hx36H89tq7uoQ6MiBx"}]}]}} {"custom_id": "Natore_10773_14877_11286_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KuMZuP2fxGLdjrDvUypeyV"}]}]}} {"custom_id": "Patuakhali_18981_9234_19494_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-63u7bpbHUdGjaV2odKhn2i"}]}]}} {"custom_id": "Bandarban_19494_48735_20007_49248.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CEBpC2CYRWUqJX3u6kR1uo"}]}]}} {"custom_id": "Naogaon_15390_14877_15903_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AeMo5tXBuYG9SoRxBpuugm"}]}]}} {"custom_id": "Rangamati_16929_47196_17442_47709.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CaHXspnKWvA9tFcv58o1w1"}]}]}} {"custom_id": "Habiganj_13338_29241_13851_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J4Aft5zeGb1cMGtXq8k3Wx"}]}]}} {"custom_id": "Jhalokati_14877_9234_15390_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Y6cftdFjQts5nf9k9YxUDg"}]}]}} {"custom_id": "Kishoreganj_13338_12825_13851_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Vw6L2MbBcYSMzhShaB4o5p"}]}]}} {"custom_id": "Nilphamari_13338_16929_13851_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FoTpXK2CW4p9KYDQBeSE1X"}]}]}} {"custom_id": "Maulvibazar_16416_9747_16929_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PWUAARv8LCw2wbZotFKjGy"}]}]}} {"custom_id": "Nilphamari_11286_6669_11799_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QkKm7KtC5ffFNDirfWZwTG"}]}]}} {"custom_id": "Chittagong_32832_22059_33345_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5wwpVzTJbz5ixys57LQtM7"}]}]}} {"custom_id": "Dinajpur_6156_12312_6669_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fad1tAb4yhq2ZyLFoRq1TD"}]}]}} {"custom_id": "Rajbari_19494_9747_20007_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KZZ59FPsqNFUFAnGric9tP"}]}]}} {"custom_id": "Natore_5643_8208_6156_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ESFYNjr6wcVB49k4yiSAyk"}]}]}} {"custom_id": "Lalmonirhat_25137_25650_25650_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SNv7KrNHJZPY8XtSnceiNu"}]}]}} {"custom_id": "Khagrachhari_16929_11799_17442_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UoigwnGTyf37SLDatWhmuj"}]}]}} {"custom_id": "Maulvibazar_13851_33858_14364_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W7yHvQfq9K4wBfufEXTbmx"}]}]}} {"custom_id": "Kushtia_22059_23085_22572_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-43gVXVkoB2X6UcWn4Bd2Pn"}]}]}} {"custom_id": "Bandarban_12825_43605_13338_44118.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CrJ8RbuWp46w454xancKvH"}]}]}} {"custom_id": "Panchagarh_11799_10773_12312_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KWUGjvu9v3bvX3ZjMLoy22"}]}]}} {"custom_id": "Maulvibazar_5130_20520_5643_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7cPzPtGDWdgw6TgjY5NtUx"}]}]}} {"custom_id": "Naogaon_15390_16416_15903_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1LH9WsKwVVWhn9zvKRhUcY"}]}]}} {"custom_id": "Noakhali_2052_1539_2565_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FnuKFPgMmqNZVnUjGFCZ3R"}]}]}} {"custom_id": "Sherpur_12312_4617_12825_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2Aq8RiPxg3FdLThwAHRmt9"}]}]}} {"custom_id": "Sherpur_14364_7182_14877_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SkSv2zFx6rgz1e1ecSf9GN"}]}]}} {"custom_id": "Thakurgaon_14364_15390_14877_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LMrrnFLGcpKrdqN1G7sAaw"}]}]}} {"custom_id": "Patuakhali_8208_34371_8721_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L7XQ3t8k8waXWjiPuip2VQ"}]}]}} {"custom_id": "Thakurgaon_4104_4617_4617_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CjKJ72XcavHjtGUgqs6Sm9"}]}]}} {"custom_id": "Jessore_11799_9234_12312_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M15Mj7kamTPrkbfrKxiWHb"}]}]}} {"custom_id": "Naogaon_22059_18468_22572_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FpenFYa1cHfZ5DgmPwREq3"}]}]}} {"custom_id": "Kishoreganj_17442_21546_17955_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M4YoP4EQUgdtjmfmS8z1Cd"}]}]}} {"custom_id": "Nawabganj_20520_6156_21033_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LujVGaztQxncAgDXyJ5wbK"}]}]}} {"custom_id": "Chandpur_18468_14877_18981_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DqrTZy9Z3SZ9xrAAoPy4ce"}]}]}} {"custom_id": "Satkhira_14364_27702_14877_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MoBx6BvEJsG6nYnP831a74"}]}]}} {"custom_id": "Khagrachhari_15390_16929_15903_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MbMapWHW5satjoFs3MobUB"}]}]}} {"custom_id": "Natore_11286_7182_11799_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T422KFGTu2MM19cosubigj"}]}]}} {"custom_id": "Lakshmipur_513_8208_1026_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GxcTbDt6wm21nHfYit4J2o"}]}]}} {"custom_id": "Lalmonirhat_14877_23085_15390_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MxvsMUMR8gsc7qbYtDeqZL"}]}]}} {"custom_id": "Mymensingh_10260_28728_10773_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KzS2PHiNY8G8uaRbrVrxtW"}]}]}} {"custom_id": "Brahamanbaria_12312_23085_12825_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JoqDhYqVNjqmrJgNgedgm2"}]}]}} {"custom_id": "Khulna_15390_28215_15903_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7cKXp7ztiqpbVnmCM5Erkf"}]}]}} {"custom_id": "Netrakona_16416_17442_16929_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AzTDYMcmVs51rkDnWxMdZj"}]}]}} {"custom_id": "Rangamati_10773_14877_11286_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WoVh28EEE1uE3FmmLAWQig"}]}]}} {"custom_id": "Sylhet_34884_7695_35397_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Rn6pH596xJo3QwJgk9TxPZ"}]}]}} {"custom_id": "Maulvibazar_12825_19494_13338_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1acAw8F4di17sRWwtR9CAg"}]}]}} {"custom_id": "Rangamati_10260_69255_10773_69768.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Rv8RzHxeFfNCy4VUf6p4N7"}]}]}} {"custom_id": "Chittagong_28728_37962_29241_38475.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8uAQEBGbPWxN2w9Q3nQGGx"}]}]}} {"custom_id": "Panchagarh_17955_31293_18468_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NHfGSJWN1rwyCUH3tDRdkT"}]}]}} {"custom_id": "Lakshmipur_14364_31806_14877_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WGUjpi7zcTrZWETAYCJer6"}]}]}} {"custom_id": "Feni_8721_24624_9234_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GvGQhT9P5d5ogwXCuL1jBj"}]}]}} {"custom_id": "Nawabganj_11799_3078_12312_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HheiGUZyJRo19Ri1oo4F1m"}]}]}} {"custom_id": "Bhola_10773_46170_11286_46683.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VBWVmeUDsuFmckqwo4pUNr"}]}]}} {"custom_id": "Natore_6156_12825_6669_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DMyXUrQ64vGbx16TWi5jn8"}]}]}} {"custom_id": "Jhenaidah_13851_10773_14364_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XkWzdGZRSWCvVobiqGd9FB"}]}]}} {"custom_id": "Chittagong_25650_15903_26163_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5HjWxcGqPHUDZvsq8avJC7"}]}]}} {"custom_id": "Natore_9234_18468_9747_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9U29r8mEYkJSBwK8M6S2o1"}]}]}} {"custom_id": "Kurigram_10773_16929_11286_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CmyELakMLBMuRakg6vrNbk"}]}]}} {"custom_id": "Narail_8208_12825_8721_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RjoFTcZKNUzRNxzq7WTqMj"}]}]}} {"custom_id": "Bogra_23598_8721_24111_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7cPe5TMNhYJKWsbSBPywmT"}]}]}} {"custom_id": "Bandarban_10260_9234_10773_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4u4HTeusMMXvUkxYxd8pGb"}]}]}} {"custom_id": "Feni_513_15390_1026_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AsPd7zmdHrTa9F4ThSVMzY"}]}]}} {"custom_id": "Mymensingh_9234_35397_9747_35910.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QwsEYpLriBqTafSt38FMsQ"}]}]}} {"custom_id": "Nilphamari_13851_6669_14364_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HxfuKxz3z2TXoANTP7SqEV"}]}]}} {"custom_id": "Dinajpur_9747_2052_10260_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-H6PdQ1HLaX4QWsVTHuLG9C"}]}]}} {"custom_id": "Sherpur_12312_11286_12825_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Wd33Gd9jJCwB3meseQcPfE"}]}]}} {"custom_id": "Sherpur_3078_16416_3591_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QWd3YcrmkoMYXSYujnuYNN"}]}]}} {"custom_id": "Pabna_25650_12825_26163_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CZ1UMQiyPpgoNT8uSwbWTj"}]}]}} {"custom_id": "Jhenaidah_28215_3591_28728_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FttyGqStzca2C69rW2ZQjV"}]}]}} {"custom_id": "Jamalpur_9747_41040_10260_41553.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8tyiHAjSe8BRVMBh6463vn"}]}]}} {"custom_id": "Gopalganj_14877_20007_15390_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Rk2WkauZsUaF2gX9cMe5zr"}]}]}} {"custom_id": "Dinajpur_24111_16929_24624_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-32RvMyHnbpTX34VsHeps2x"}]}]}} {"custom_id": "Chittagong_21033_15903_21546_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Pahz96GY5xxi4pg7ehtZeh"}]}]}} {"custom_id": "Rangamati_9747_2565_10260_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UPBNwrVCrE84QxpuzxGekJ"}]}]}} {"custom_id": "Rangamati_27702_83106_28215_83619.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7uSkStrzUDxu5V3zQBW2Fb"}]}]}} {"custom_id": "Netrakona_23598_16416_24111_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Y4n5qYrLaTxntD13Jbg1fj"}]}]}} {"custom_id": "Jessore_8208_513_8721_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3ssUrGpNs1UCAZY7Hd4wLn"}]}]}} {"custom_id": "Rajshahi_14877_5130_15390_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8zatBwa8fL78ZYuCKBB4TK"}]}]}} {"custom_id": "Maulvibazar_3591_30780_4104_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2iQtEvnKE35jHvQdnJEMKj"}]}]}} {"custom_id": "Mymensingh_21546_45144_22059_45657.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5sUtyNtnKCFjxCDx6FjqXv"}]}]}} {"custom_id": "Rangamati_10260_12825_10773_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K7ogCcVvXX9Q7NAzdz5bjM"}]}]}} {"custom_id": "Tangail_7182_33345_7695_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W3Sc2v5c5pUJqJjFmKZpZx"}]}]}} {"custom_id": "Habiganj_7695_30267_8208_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4TP85iQVsHwcs4k4PKeb1j"}]}]}} {"custom_id": "Sunamganj_23598_11799_24111_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1frDNUGBUc2UhYVKZ3o4JZ"}]}]}} {"custom_id": "Kishoreganj_7182_7695_7695_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6bmrsWMcPWWy7nVo7E4u97"}]}]}} {"custom_id": "Thakurgaon_9234_2052_9747_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2FDruNPsy8Po4ovDhfjEFw"}]}]}} {"custom_id": "Bagerhat_14364_26163_14877_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fj21D9NrrkSBRccFT9zWmJ"}]}]}} {"custom_id": "Kishoreganj_16929_1539_17442_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NTXTJi4KFetuvWbsPRcxmT"}]}]}} {"custom_id": "Nilphamari_13851_12825_14364_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-A448Rnnr4mnX9dMnPuztof"}]}]}} {"custom_id": "Chittagong_35397_20520_35910_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LZmgaGvsP1Lc2w2DvrZbhr"}]}]}} {"custom_id": "Jessore_6669_8721_7182_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4KkcvZfmDQyDoT7vL8Ud3B"}]}]}} {"custom_id": "Chittagong_11286_27189_11799_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4bsW9QJZVdz6TarG4by4aU"}]}]}} {"custom_id": "Narayanganj_9234_1539_9747_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Pc8m8o92dijtaFvUcSR5sS"}]}]}} {"custom_id": "Rangamati_12825_17955_13338_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DrcQTy2uum1qpiPFVALWz7"}]}]}} {"custom_id": "Pirojpur_3078_4104_3591_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NmMUvr5nGGs7ZUHZ79SmCf"}]}]}} {"custom_id": "Bagerhat_5130_27702_5643_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XZ4y2Yv3fvp4gsNZjLs68s"}]}]}} {"custom_id": "Barguna_7695_10773_8208_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FoZ1CPCZt6bMcMoHrerjQC"}]}]}} {"custom_id": "Sunamganj_21546_2565_22059_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3qsb8cKXwasSQ2ic2PSJ98"}]}]}} {"custom_id": "Chandpur_12312_13338_12825_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1Ln5aX2hS7gzwPxrMPKhLF"}]}]}} {"custom_id": "Barguna_4617_26676_5130_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hm7KSbsSHJnw7do54MVdex"}]}]}} {"custom_id": "Chittagong_31293_42579_31806_43092.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FQPP5VRcPN28oSGCtZc69c"}]}]}} {"custom_id": "Sherpur_11286_20520_11799_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-34AcdhyZLXqKj5yVKmTUe2"}]}]}} {"custom_id": "Rangpur_12825_1026_13338_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KHrTzTVYJj7Z87xfcZ9jRL"}]}]}} {"custom_id": "Comilla_23598_23085_24111_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7zMprM7MoFfxXKKvhwGdbf"}]}]}} {"custom_id": "Rangpur_8208_26163_8721_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-52Y7wRTYBom5cp6J3ZjT7d"}]}]}} {"custom_id": "Lalmonirhat_7695_4617_8208_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-25kzEnQDAE2qA9xsWK1LDh"}]}]}} {"custom_id": "Bhola_6669_28728_7182_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1PxvXHWAcJnRkr1bUc1RHx"}]}]}} {"custom_id": "Naogaon_23085_23598_23598_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7awC6XzxYSGbdxRS72wdfM"}]}]}} {"custom_id": "Sylhet_25137_12825_25650_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mt8KJiwYDTCYS2xy6jkcKU"}]}]}} {"custom_id": "Bandarban_10773_15903_11286_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VQgst9zENnrCaTBbtyYzdk"}]}]}} {"custom_id": "Maulvibazar_14364_33858_14877_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R5anSoRBRU9QxgANoMbhxD"}]}]}} {"custom_id": "Feni_8721_15903_9234_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VN1VDHXyoNccdpMrJt9Yg3"}]}]}} {"custom_id": "Bogra_15903_19494_16416_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Q1tJrv6GQmPc7ZEgNLz2pd"}]}]}} {"custom_id": "Habiganj_12312_24111_12825_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R3uB94mcLg6EGWZ2gw5iTE"}]}]}} {"custom_id": "Mymensingh_23598_23085_24111_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BcTK7ZYFZGK473LntqiCvi"}]}]}} {"custom_id": "Nawabganj_4104_17955_4617_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VrY9AJqBwZoEJLZqt1uvQH"}]}]}} {"custom_id": "Sunamganj_15390_22059_15903_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Y467DAyBEoRob9tb8MQhYP"}]}]}} {"custom_id": "Nilphamari_10260_9234_10773_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AS7eNh4u1XvoxxdyoqvkeC"}]}]}} {"custom_id": "Rangpur_13851_6156_14364_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-731abyRotVLsCsAa5v3GwG"}]}]}} {"custom_id": "Nilphamari_15390_21033_15903_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3Xhit8P2y1Lm4SHmM3nfzy"}]}]}} {"custom_id": "Mymensingh_16416_35910_16929_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KxyRSQuwjUepSKSQY3tLcT"}]}]}} {"custom_id": "Jessore_23598_18468_24111_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Dpbwr5QqNqPxHYtM226xpX"}]}]}} {"custom_id": "Tangail_5130_37962_5643_38475.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U44HxUmjiK2zX7VuvPHxVQ"}]}]}} {"custom_id": "Noakhali_16929_20520_17442_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SaprMGPMLG9q69dpcGVguu"}]}]}} {"custom_id": "Sherpur_8208_17442_8721_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Po9Gzih3uPfjPJEfHB4KKW"}]}]}} {"custom_id": "Mymensingh_11799_36423_12312_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ABAKXNSwUiYzeT936KCr89"}]}]}} {"custom_id": "Chittagong_16929_3078_17442_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Kq2VyKb6Cno98gaiBioUo6"}]}]}} {"custom_id": "Naogaon_8208_14877_8721_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5NnchH98FyRq157cv9p73x"}]}]}} {"custom_id": "Gopalganj_9747_24111_10260_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SatoLmUUDf3GkqRYMdPu4p"}]}]}} {"custom_id": "Bandarban_6669_30267_7182_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UX2V3SqKVv88JTkYquiv7V"}]}]}} {"custom_id": "Sunamganj_26676_23085_27189_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TKwp9pTq4881pjbmFF5aMu"}]}]}} {"custom_id": "Feni_8721_25137_9234_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JCF1wmHghGdzVUeMXcQkhg"}]}]}} {"custom_id": "Habiganj_7182_25137_7695_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JFuw1W2N6suRBy4i3zVTCh"}]}]}} {"custom_id": "Bogra_27189_22059_27702_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BRVeUGijvmr5ZrGbWNWr5L"}]}]}} {"custom_id": "Sirajganj_17442_31293_17955_31806.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RPGRf59PgzSWC34q2bHCLp"}]}]}} {"custom_id": "Rangamati_19494_18981_20007_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AmTAZXKVPrk5hibCzJf9Zc"}]}]}} {"custom_id": "Thakurgaon_1539_10773_2052_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MbPBw1SXv83948VfU7Mt34"}]}]}} {"custom_id": "Chittagong_14877_8208_15390_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V9mQygwC4ce6jnbHJN76nU"}]}]}} {"custom_id": "Satkhira_6156_5130_6669_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UNdmo6CVuASfosthgvDhNM"}]}]}} {"custom_id": "Rangamati_12312_33345_12825_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XVfBQ4cdBQx3E4BSy4Hbuc"}]}]}} {"custom_id": "Sirajganj_21033_6156_21546_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gpj6LgUaCdobTNRrgzUogV"}]}]}} {"custom_id": "Dinajpur_3591_10773_4104_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hm1WzeNRJmpGHDv8RwW6ub"}]}]}} {"custom_id": "Gaibandha_15390_4617_15903_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8dxoY3JNfhQGcDRmCadFt4"}]}]}} {"custom_id": "Narail_11286_8721_11799_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4ZAsZCjwiK1ydWLeRwBqHj"}]}]}} {"custom_id": "Nawabganj_9234_24624_9747_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NkHvVeXo9k29yVjT3Ye9fi"}]}]}} {"custom_id": "Jessore_13338_7695_13851_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8YsuffnBvE7nG2YbzxZFTs"}]}]}} {"custom_id": "Gazipur_13851_7695_14364_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BREaMmzAoBdogWRiFpW5NL"}]}]}} {"custom_id": "Sunamganj_16416_30267_16929_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7kJM2kU3MfMkneqWv8ec9N"}]}]}} {"custom_id": "Kishoreganj_14364_13338_14877_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MiFg13aAWdGKNBMNtMrP57"}]}]}} {"custom_id": "Rangamati_14877_50274_15390_50787.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GMkKqibBv1Vj3m11qFtoUw"}]}]}} {"custom_id": "Mymensingh_6156_26676_6669_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-P2fJDWj3em2ijBCE4zBwKK"}]}]}} {"custom_id": "Jessore_16929_17442_17442_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BmTdXmCcRdyVWXUG4HBoWv"}]}]}} {"custom_id": "Rajshahi_26676_30267_27189_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7gszoFubHZ2MV8APHJufbk"}]}]}} {"custom_id": "Maulvibazar_2052_30780_2565_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NkwRdjbvGiBkU6rxi9uPuw"}]}]}} {"custom_id": "Narayanganj_2565_13851_3078_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TAtWoQxPggdS6aLkrkoLva"}]}]}} {"custom_id": "Kushtia_5643_6669_6156_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8RCCYRLssYmqBazCVPnVRy"}]}]}} {"custom_id": "Sirajganj_24111_2052_24624_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MX6Gqi6Kuac9QPCtyU6fPB"}]}]}} {"custom_id": "Satkhira_5130_14364_5643_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AUrqeQKifCWSkQbwDirfcS"}]}]}} {"custom_id": "Sylhet_7182_12825_7695_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LVGwJ3UyJ2PUETmFCUZWpS"}]}]}} {"custom_id": "Natore_10260_7695_10773_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SWk6tN3hpHNMTF4BEBpxfK"}]}]}} {"custom_id": "Pabna_14364_13851_14877_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-25aFA7xAMo3v6TZTvdzyTQ"}]}]}} {"custom_id": "Comilla_15390_13851_15903_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XS6FWDRStThh7a3uE6mGKz"}]}]}} {"custom_id": "Sunamganj_24624_6156_25137_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mdo95rff3XRtatLEcwLr1P"}]}]}} {"custom_id": "Maulvibazar_6669_21033_7182_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MDm7KXU4sURrCqyf5xQ6ZC"}]}]}} {"custom_id": "Narail_13851_17442_14364_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W5hgRHYLQsnag4DwFHHJvc"}]}]}} {"custom_id": "Rangamati_13851_65664_14364_66177.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SXQux976bvDwNh9NyBWYMD"}]}]}} {"custom_id": "Satkhira_14877_16416_15390_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-E5hjGV26aYD2dnnV2kryzF"}]}]}} {"custom_id": "Panchagarh_9747_22059_10260_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Pg9CcoDZeJbMM3F7DLyZpD"}]}]}} {"custom_id": "Nawabganj_16416_13338_16929_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LBXifXrKB96VkwftkuZCgy"}]}]}} {"custom_id": "Narail_17955_6669_18468_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1sDMrAVVUdMPjsmt83Ydtw"}]}]}} {"custom_id": "Shariatpur_13338_3591_13851_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GuwU8rPcUMSrxcjNwEucN1"}]}]}} {"custom_id": "Bogra_8721_20007_9234_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HD68BbeWiwZX4thxCNJ5G8"}]}]}} {"custom_id": "Feni_5643_24624_6156_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9r1ynpza8ZEWMJs5tAghxQ"}]}]}} {"custom_id": "Dinajpur_15390_6669_15903_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KqPHkdyMDyCsXLbHaATDhH"}]}]}} {"custom_id": "Khagrachhari_4104_41040_4617_41553.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XJXxvDNFzdRgaeTPb6PtwB"}]}]}} {"custom_id": "Habiganj_18468_8721_18981_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AnrANcCdpn5gd2E95XYNvL"}]}]}} {"custom_id": "Kurigram_14877_12312_15390_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AVCS9ogNCzU1J2VBeZC9px"}]}]}} {"custom_id": "Gazipur_7695_12825_8208_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4zHZK4mzSYGokgkzhuuDHP"}]}]}} {"custom_id": "Brahamanbaria_16929_1539_17442_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3VoDY5mUMqb5XnUME7VCq5"}]}]}} {"custom_id": "Khulna_11286_53352_11799_53865.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QSQvxpqkQ9tZ64UhFgsbd5"}]}]}} {"custom_id": "Mymensingh_7182_27189_7695_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mtko9aE19SmxEcsHafkz4y"}]}]}} {"custom_id": "Satkhira_9234_20520_9747_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6sMR8rhKC1oBNk8sAA2Q4q"}]}]}} {"custom_id": "Bagerhat_13851_28215_14364_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BW2UzfE219kbJGx4Ccr9dH"}]}]}} {"custom_id": "Mymensingh_26163_40014_26676_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QiHUw7Rt1zQogKuZtUKgk2"}]}]}} {"custom_id": "Khagrachhari_12312_10773_12825_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1azvj6UJuzuB7KoiXehznT"}]}]}} {"custom_id": "Nilphamari_2565_9747_3078_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RLkjaPW1nKfbvrK9rkCsvV"}]}]}} {"custom_id": "Maulvibazar_7695_28215_8208_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YSmRUMoNNqbomvcjuks2F2"}]}]}} {"custom_id": "Gazipur_13338_21033_13851_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RALvehyd3xXRjUBhcWXLav"}]}]}} {"custom_id": "Mymensingh_20520_2565_21033_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NSRxnBL1VYnfZmpHVHS8BU"}]}]}} {"custom_id": "Pirojpur_4617_25137_5130_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WcS3vpBDd2sWby2qsu44Nm"}]}]}} {"custom_id": "Chittagong_30780_32319_31293_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4kgQRugibXVgzpeggHcBta"}]}]}} {"custom_id": "Gazipur_11799_13338_12312_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RJFi5tLhTtU5m7r2Pgtq1j"}]}]}} {"custom_id": "Rangpur_21546_10260_22059_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-39yGx5gDqhjB8hjrMM9ThG"}]}]}} {"custom_id": "Jessore_24111_10773_24624_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6yztCtYbR5f3ux3ZjxgbrK"}]}]}} {"custom_id": "Chittagong_24624_32319_25137_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RMfrVqayWnKd7gxKcFBWTr"}]}]}} {"custom_id": "Bagerhat_13338_55917_13851_56430.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6QN6ajiocMVZJRonMY2mds"}]}]}} {"custom_id": "Rangamati_11799_11286_12312_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8Len6cz5ZjWSRNwGRQEKtJ"}]}]}} {"custom_id": "Satkhira_12312_33858_12825_34371.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JbcNRGCBzvUFbADcKsBknc"}]}]}} {"custom_id": "Netrakona_22059_27702_22572_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GbovUnW2kWuDhmYRuW7s4N"}]}]}} {"custom_id": "Tangail_14364_36936_14877_37449.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4LF482BwS6gtn7JLV3tqVA"}]}]}} {"custom_id": "Bagerhat_4104_45144_4617_45657.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BpYs2K9Gr8CtksrNjy3CzX"}]}]}} {"custom_id": "Panchagarh_7182_4617_7695_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U68b6mATYpb1X8tu8966nH"}]}]}} {"custom_id": "Pabna_21546_22059_22059_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JezpzJAZtuK8JJtvLEv9HH"}]}]}} {"custom_id": "Barguna_5130_20520_5643_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HrUjgHxhwcCDry3jZEF6pV"}]}]}} {"custom_id": "Noakhali_20520_20007_21033_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JoUHDKZwWecDEHPBssweqH"}]}]}} {"custom_id": "Faridpur_28215_14877_28728_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YFciLzsumLn4hvzNLa4mDa"}]}]}} {"custom_id": "Netrakona_25650_23085_26163_23598.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BSmSxz1j2PVYXKJeVYZ1Pg"}]}]}} {"custom_id": "Jessore_13851_11286_14364_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1ufmKBBdh8chBHWREJqJNL"}]}]}} {"custom_id": "Jamalpur_9234_10260_9747_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9XgagCQXoD4bzWbwZvTfJV"}]}]}} {"custom_id": "Sherpur_10773_17442_11286_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KPtAjEtYWKBwhiQsEu2oa3"}]}]}} {"custom_id": "Sunamganj_18468_15903_18981_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MiESjC4J2bRqGeGTysGMPs"}]}]}} {"custom_id": "Comilla_22059_16929_22572_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YQMZopJn7B68JqHDyCzt1k"}]}]}} {"custom_id": "Rangamati_6156_40527_6669_41040.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WrfVMpYA1GvbXrQKKqVMc6"}]}]}} {"custom_id": "Rangamati_31293_88236_31806_88749.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8d5WAp8spbxTjgW3pXppgF"}]}]}} {"custom_id": "Narail_10260_13851_10773_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-77MLWFSr3Zyz5tnu16Ni9L"}]}]}} {"custom_id": "Bagerhat_12312_6156_12825_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8EqFGUpX1WJRL55ZphxLQY"}]}]}} {"custom_id": "Gaibandha_18468_14364_18981_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CcSzXUnB4MVGGRtu57SzwM"}]}]}} {"custom_id": "Rajshahi_13851_8208_14364_8721.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Y9bHE5hJy1fU38ELh4TG7q"}]}]}} {"custom_id": "Netrakona_11799_8721_12312_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6iTPQYv4TbSXyfXpKzm62a"}]}]}} {"custom_id": "Mymensingh_21546_29241_22059_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Y9VzbHBvkPx7pu8tp874Xy"}]}]}} {"custom_id": "Barguna_11799_26163_12312_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VMB2K1D56RatJqPrCp9rPw"}]}]}} {"custom_id": "Bandarban_5643_15903_6156_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Nw1FJN8hLo3dEwfLQKH8cX"}]}]}} {"custom_id": "Narail_7182_11799_7695_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RDvqGn3keVGDmGr2uoZn9g"}]}]}} {"custom_id": "Rangamati_10260_54891_10773_55404.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MHhLXQrzggo5oR9Yx77RoY"}]}]}} {"custom_id": "Mymensingh_29241_34371_29754_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UZxKFVGtiCKbcGQSKUjW1L"}]}]}} {"custom_id": "Chittagong_27189_47196_27702_47709.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RKuBAbJfJ6AqD9jtjsfmQq"}]}]}} {"custom_id": "Sylhet_8721_23598_9234_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YV6qGX4A2R7dKBGM69frVi"}]}]}} {"custom_id": "Kurigram_18468_30267_18981_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L5BY3oUCPRqJFs6gPRKSxB"}]}]}} {"custom_id": "Noakhali_9234_26163_9747_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7nbdvJ3TANtJns3zcMchfM"}]}]}} {"custom_id": "Rangpur_9747_24111_10260_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TnEMGuUGXnBkcgySs5R4wA"}]}]}} {"custom_id": "Khagrachhari_4104_45657_4617_46170.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TfkZZ6kHv7BNUEasB34JAA"}]}]}} {"custom_id": "Narsingdi_4104_9234_4617_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NmKodhYYExh1qnzbFXbiTx"}]}]}} {"custom_id": "Barguna_4104_13338_4617_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9Q1NvGmm1p1Tc9o3QStqoL"}]}]}} {"custom_id": "Lakshmipur_2052_5130_2565_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9noq1VA7MZXsATNYV82eyp"}]}]}} {"custom_id": "Jhenaidah_9747_26163_10260_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6rdny9fPFmLuWY9cKMUyVu"}]}]}} {"custom_id": "Bagerhat_9234_34371_9747_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VxXRdTq3Kt5yZB8qetkU66"}]}]}} {"custom_id": "Sylhet_38475_14877_38988_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4BjPcBHwDhuhMxnzUPYVSA"}]}]}} {"custom_id": "Sylhet_17442_7695_17955_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-34kFtcywR8D6dmMgL1G7RC"}]}]}} {"custom_id": "Rangamati_16929_42066_17442_42579.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5murWhf85ryxKmfP9ksgM4"}]}]}} {"custom_id": "Bagerhat_13851_28728_14364_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-98csrZ6gt2yRJsVHBm7gH4"}]}]}} {"custom_id": "Bhola_9747_18981_10260_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Y8o4SrtPwnDntbqocaZJkg"}]}]}} {"custom_id": "Lakshmipur_6669_13338_7182_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SzL51h4y99CRc3LxBv3eje"}]}]}} {"custom_id": "Khagrachhari_15903_14364_16416_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4AATynrZgApqPC3m95K3Ns"}]}]}} {"custom_id": "Bandarban_17442_4617_17955_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N8iFWpZaLKkcUdS5RR31Q9"}]}]}} {"custom_id": "Feni_14364_15903_14877_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TAFMTMWLgjjtMxQNxFDAL1"}]}]}} {"custom_id": "Bagerhat_7182_32832_7695_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WU5u5McLjAAjbV78PWmjuF"}]}]}} {"custom_id": "Naogaon_20007_9234_20520_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2aE6zw7WH9AbbsUCyrpKpH"}]}]}} {"custom_id": "Nawabganj_5130_11799_5643_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AULVeqG3v5imzqjSdD48sU"}]}]}} {"custom_id": "Thakurgaon_17442_3591_17955_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Caye6XQv28LvPU8PJEswVH"}]}]}} {"custom_id": "Nawabganj_12825_16416_13338_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WGnY1Hv3JSu9rTAD2pvASs"}]}]}} {"custom_id": "Madaripur_16929_20007_17442_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RfhVDQncHqCuCC16HKZuWL"}]}]}} {"custom_id": "Bandarban_17442_48222_17955_48735.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5xqGGD7SZkv6p1tXVK4FNG"}]}]}} {"custom_id": "Maulvibazar_2052_25137_2565_25650.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6u9D8Fajfsjn1nX3oVfYhE"}]}]}} {"custom_id": "Noakhali_21033_30780_21546_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N7xj3cfjrpABdriCikc5QR"}]}]}} {"custom_id": "Feni_513_20007_1026_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FKgfCNooydQHYZH9CWZd8F"}]}]}} {"custom_id": "Rangpur_27189_14877_27702_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VABz594RbmkPcy8BwXH4vQ"}]}]}} {"custom_id": "Dinajpur_11799_19494_12312_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3XxtQqSH9gqavJcYbdnekm"}]}]}} {"custom_id": "Sunamganj_35397_13338_35910_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KoGAFX4NRxQAN8u4Qt3C4t"}]}]}} {"custom_id": "Nawabganj_10260_1026_10773_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EwsGCRGQEQtBSJjZkSffLQ"}]}]}} {"custom_id": "Sirajganj_20520_32319_21033_32832.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-YY7UUfUq8y2B2TfieatA7e"}]}]}} {"custom_id": "Naogaon_20007_11286_20520_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WGNztXUyKcd58V1232A6W3"}]}]}} {"custom_id": "Feni_13851_12312_14364_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PPpkoZbugRbiJ777Jzu3AX"}]}]}} {"custom_id": "Chittagong_40014_50787_40527_51300.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VMQ8pATMqXuGSf5dhr6Epg"}]}]}} {"custom_id": "Mymensingh_25137_24111_25650_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BbiMcb2X7n2qU1jqVnnegj"}]}]}} {"custom_id": "Lakshmipur_8721_14877_9234_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7bKoVyCzScH1m2zemTLaWS"}]}]}} {"custom_id": "Patuakhali_8208_7695_8721_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ef6APS2VzLc4oeJMgzCwxD"}]}]}} {"custom_id": "Bandarban_6156_8721_6669_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5MPbmRLUVcFWPSBz7wD3Rd"}]}]}} {"custom_id": "Rangpur_20007_21546_20520_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JDYUC6zVdNpm4Yg7A3KPc2"}]}]}} {"custom_id": "Bhola_3078_40014_3591_40527.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RFk4Rdjgw6E91R54j5YXmN"}]}]}} {"custom_id": "Gaibandha_3591_26676_4104_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UqXeEwWLyhMZmZBd8T9shM"}]}]}} {"custom_id": "Sirajganj_3591_13338_4104_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hr5io2PYjJn1FVYaz5cRow"}]}]}} {"custom_id": "Kurigram_5643_21546_6156_22059.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RM1RxETayYBroFTyWtTSi4"}]}]}} {"custom_id": "Narsingdi_7695_10260_8208_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-57GGjUFAKDeLSK5NK5t1S4"}]}]}} {"custom_id": "Comilla_24111_7695_24624_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XjAVrggLpaFbDQGgMrwEXd"}]}]}} {"custom_id": "Chittagong_23085_37962_23598_38475.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LEbyCMyAV5Q8oBrzWMVU88"}]}]}} {"custom_id": "Patuakhali_16416_28728_16929_29241.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-76vViv4yB4uMbiDnHnogAL"}]}]}} {"custom_id": "Sylhet_15390_14364_15903_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Wwh4A83izEXwtZi2zfvRZF"}]}]}} {"custom_id": "Netrakona_24111_15390_24624_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KWLV8JEvV2ZqHsreNDZZHq"}]}]}} {"custom_id": "Narsingdi_8721_14877_9234_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6uPJxcUtxin1sJbzAbmLAa"}]}]}} {"custom_id": "Gazipur_18468_9234_18981_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GWhp8uV1wQ8LAg9iEEn4LW"}]}]}} {"custom_id": "Tangail_13338_14877_13851_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CatrFoQ1tHzTBwo7dogiTg"}]}]}} {"custom_id": "Habiganj_5130_5130_5643_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NxT3t2qZpmA39CL4ZsgA1m"}]}]}} {"custom_id": "Kishoreganj_18981_28215_19494_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RnYhBhTsp2sPXncDFi7u5p"}]}]}} {"custom_id": "Barguna_8721_14364_9234_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6xffUaaCmhrWL51wWQZDND"}]}]}} {"custom_id": "Patuakhali_16929_24111_17442_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GDyUgw857WQF4wnZc5h8gY"}]}]}} {"custom_id": "Barisal_11799_5130_12312_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6RpYaPzceVNX3Ki3iBjdGv"}]}]}} {"custom_id": "Khulna_10773_9234_11286_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2X9RcWhHx1dm5TEEd6g2Lw"}]}]}} {"custom_id": "Sherpur_6156_3591_6669_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4s2YPCbwzMzBc2PbSjfkZH"}]}]}} {"custom_id": "Barisal_20520_21033_21033_21546.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2jxBsfthzE9ztzPuDQQGAG"}]}]}} {"custom_id": "Tangail_6156_27702_6669_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KJTAceBTWAi5ZxDGyfxaXZ"}]}]}} {"custom_id": "Gopalganj_6156_9234_6669_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V54tQ9o7iHjaDekHhFQ1Hp"}]}]}} {"custom_id": "Pabna_22572_7695_23085_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8FfvAanUQa4NhpvYspDRak"}]}]}} {"custom_id": "Chuadanga_14877_5130_15390_5643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Q2LHwBdnq7AZwCD8aZnbps"}]}]}} {"custom_id": "Jamalpur_20520_30780_21033_31293.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Lp5n8BPwXjJJbVmAg5jUkp"}]}]}} {"custom_id": "Tangail_14364_26676_14877_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U3CSyoEqms7ByCKRuG3oiz"}]}]}} {"custom_id": "Habiganj_7182_29241_7695_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DjwJVLspGLX7udSZ8wMoim"}]}]}} {"custom_id": "Rajshahi_30780_7695_31293_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-SWnkn9cmgE3c9wGCeCp8Jr"}]}]}} {"custom_id": "Satkhira_16416_16929_16929_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JPYDo2p1PDgyk3LA6qGFBA"}]}]}} {"custom_id": "Bagerhat_14877_41040_15390_41553.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-PNYzk1Vj8sbtDatcjJPpG4"}]}]}} {"custom_id": "Netrakona_15903_29754_16416_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DRW4omNRC3rerHrTj5ffAR"}]}]}} {"custom_id": "Khagrachhari_3078_42579_3591_43092.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-G2mdtQVjdiXmACrgqVXr6p"}]}]}} {"custom_id": "Jhenaidah_11799_23598_12312_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3uXPRt3T6G1zXFp5rvWhjB"}]}]}} {"custom_id": "Patuakhali_22572_34884_23085_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8w19Xe6TfDKFQBL6AvmQg9"}]}]}} {"custom_id": "Jessore_20007_13338_20520_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NvZqwSwqqaTVSmqzicPG69"}]}]}} {"custom_id": "Brahamanbaria_12825_22572_13338_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3Vsfwtjmdgi1nFEz54ZUjv"}]}]}} {"custom_id": "Satkhira_12825_20007_13338_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4Re1HZEXEk83EsXFyMmUoE"}]}]}} {"custom_id": "Narayanganj_10773_8721_11286_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-G4TmqeKUdTpEAqEHTuqVdN"}]}]}} {"custom_id": "Naogaon_11286_9747_11799_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EiCdZVDNYs7weYYCrZfrTQ"}]}]}} {"custom_id": "Sirajganj_23085_10260_23598_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VbiVCNHztW6GMVMT1CTj4h"}]}]}} {"custom_id": "Pabna_11286_6156_11799_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BZADgiGQ77MsNqNjFpVv9i"}]}]}} {"custom_id": "Dinajpur_23598_35910_24111_36423.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BeYusN5p9nLjKsy7owJtEk"}]}]}} {"custom_id": "Gazipur_20520_6156_21033_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-S7aqKt2wLFxKgp6FHAbwyz"}]}]}} {"custom_id": "CoxsBazar_9747_34884_10260_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-731RqDPAkyBpvs41jQ5WYG"}]}]}} {"custom_id": "Magura_10773_14877_11286_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5tnB7UY1KSAcsyHbNSdmaN"}]}]}} {"custom_id": "Habiganj_12825_27189_13338_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-83J6Lq63BSz2GDbJkZSNmR"}]}]}} {"custom_id": "Tangail_1539_16929_2052_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1aKHmTGzcuW1ctoXcJ9wCa"}]}]}} {"custom_id": "Sylhet_15903_11799_16416_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KN7iHrPUD5DddaWRSLVAYL"}]}]}} {"custom_id": "Sirajganj_17442_6669_17955_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-37kxYpRundtYhXn4nn56io"}]}]}} {"custom_id": "Nawabganj_17442_1539_17955_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W1LSHzDCThVySyDyZeB1Uk"}]}]}} {"custom_id": "Bogra_18468_24111_18981_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CvfwQqv1x3ssW5vyheiSns"}]}]}} {"custom_id": "Nilphamari_3078_16929_3591_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1c62AFpBnzv5bLfrhHkVKF"}]}]}} {"custom_id": "Sunamganj_30780_12312_31293_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VNt4giB6QqePNsB3zkQ5cb"}]}]}} {"custom_id": "Panchagarh_8721_14364_9234_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7uEMcKpgLEdmkaAM9iRLFS"}]}]}} {"custom_id": "Dinajpur_10773_28215_11286_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-T2ZiQanKp6zRJ8JzzWMtZH"}]}]}} {"custom_id": "Nawabganj_12312_7695_12825_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D5S41eEieGFQSc15Dr3ySg"}]}]}} {"custom_id": "Brahamanbaria_21033_2565_21546_3078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9cwvetSmKBAbt5LUnTuyiC"}]}]}} {"custom_id": "Netrakona_30267_27189_30780_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-C2Unb1bodYkVjZBrtaDHC1"}]}]}} {"custom_id": "Tangail_10260_9234_10773_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-U8PtWDcML53Gomhqd8zEdv"}]}]}} {"custom_id": "Rajshahi_15903_16416_16416_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-79DXr7J3NwDAHvEVcC1Wh8"}]}]}} {"custom_id": "Kishoreganj_26163_7182_26676_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KuUrpLXLZipMr7YsLB2MPq"}]}]}} {"custom_id": "Sylhet_24624_10260_25137_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ly7rdCP3gY81cBKnEDC2s4"}]}]}} {"custom_id": "Bagerhat_14877_22059_15390_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hj2fwFdc6rfg2Qz9ugdqHj"}]}]}} {"custom_id": "Bandarban_20520_16929_21033_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HKEcw7NizSB6oaeoVzvhgC"}]}]}} {"custom_id": "Satkhira_10773_32832_11286_33345.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-TDh2kKqJpBaQhwNGH5cy9M"}]}]}} {"custom_id": "Habiganj_3078_22572_3591_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Mmhpj58faR9AMWNPcVxNrj"}]}]}} {"custom_id": "Jhalokati_5130_16416_5643_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ViR7pA89vhviEV2fVpsJbN"}]}]}} {"custom_id": "Barisal_22572_6156_23085_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MfqzSZrbjaq9GcHq49jXTh"}]}]}} {"custom_id": "Netrakona_26676_26676_27189_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3nmpYQsxEf3ZVcBkbjQeC3"}]}]}} {"custom_id": "Rangpur_11799_9234_12312_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Cpc5S12ybJ68ZcBWvgcPD1"}]}]}} {"custom_id": "Barisal_24624_3078_25137_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-N2n7wgCxZLpGkyzGptggE3"}]}]}} {"custom_id": "Rajbari_11286_9234_11799_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9DXNMjATrEqFpMeoY1iikB"}]}]}} {"custom_id": "Mymensingh_23598_31806_24111_32319.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JXLXbsfMyn9ZFCthrsLvSe"}]}]}} {"custom_id": "Netrakona_13851_26163_14364_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BrobMHkb3XC4kCv37pXGDK"}]}]}} {"custom_id": "Maulvibazar_4617_19494_5130_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UghSLf7bJ4SmCTNWUQ1LRx"}]}]}} {"custom_id": "Faridpur_17442_513_17955_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-75E37F9ceVidP6esyxrgAT"}]}]}} {"custom_id": "Gaibandha_11286_27702_11799_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DjRBnhs4Jyxn3URZVrqWmf"}]}]}} {"custom_id": "Tangail_15390_7182_15903_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UmHSCxcs1NckfrUbhWTtFS"}]}]}} {"custom_id": "Patuakhali_3078_16929_3591_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4Hq9La1rThzZzB2zeWUmkh"}]}]}} {"custom_id": "Faridpur_13851_3591_14364_4104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QxCpeXXWhVw19nziLdheC2"}]}]}} {"custom_id": "Mymensingh_25137_23598_25650_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-B6x6r75DLLhaiKn7ethHmC"}]}]}} {"custom_id": "Thakurgaon_8721_20007_9234_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BGrmGrhZZmbXZSToeMLq8A"}]}]}} {"custom_id": "Patuakhali_18981_3078_19494_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RfCpXeEoHUr2vXw9jpG9zd"}]}]}} {"custom_id": "Pirojpur_4617_15903_5130_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FPxPTQPw6s8hPi8QU4Q7qg"}]}]}} {"custom_id": "Satkhira_8208_51300_8721_51813.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2vRWCTDMePoqTncZz4pnkR"}]}]}} {"custom_id": "Satkhira_7695_58995_8208_59508.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FyQsWBkNRMMkGcbpzSZpwY"}]}]}} {"custom_id": "Mymensingh_22059_6156_22572_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EGF2oVEScHAmoZHzCSkXKA"}]}]}} {"custom_id": "Khulna_17955_11799_18468_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WKxweM2nKxPABe44uafw94"}]}]}} {"custom_id": "Brahamanbaria_22572_10773_23085_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-J7C4AczPDHKDFY1ZMNSn8T"}]}]}} {"custom_id": "Netrakona_16929_27702_17442_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JrrhTzzhE4mw2BmJG64zCH"}]}]}} {"custom_id": "Jessore_8721_16416_9234_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DWEmUnkgfFoiLQtqfVfCDe"}]}]}} {"custom_id": "Chuadanga_5130_8721_5643_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QvGSbxpMqWLV4xfnLYYnXZ"}]}]}} {"custom_id": "Gaibandha_7695_16416_8208_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ACkauXwofsFRKFpYxZSxgf"}]}]}} {"custom_id": "Maulvibazar_14877_29754_15390_30267.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-R4mhN8MYUridW1XRGjrkTp"}]}]}} {"custom_id": "CoxsBazar_13338_30267_13851_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NqFJ2gUTPGU9obPYCbhhy6"}]}]}} {"custom_id": "Natore_17955_19494_18468_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7iDuZBN2tSYvzWJaXwk4Q5"}]}]}} {"custom_id": "Faridpur_23085_8721_23598_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-UpwovyP5SHFKxiKg8KAigG"}]}]}} {"custom_id": "Chuadanga_8208_14364_8721_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EyLB3twQtpCbQp4kxcdKHS"}]}]}} {"custom_id": "Maulvibazar_25650_14364_26163_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GXaZRNj2qC3vQNpDFo3jre"}]}]}} {"custom_id": "Sunamganj_26676_17955_27189_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MaWT56aWJ6zHVaRJPvvGf9"}]}]}} {"custom_id": "Maulvibazar_11286_18981_11799_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DLodonZYkKBXjsGFwYiXta"}]}]}} {"custom_id": "Kushtia_20007_20007_20520_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Wncd9tzixEdxj6JtX46RGp"}]}]}} {"custom_id": "Gopalganj_4617_6156_5130_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MC12jFumq48Y8bBRHtpGkv"}]}]}} {"custom_id": "Naogaon_25137_6669_25650_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JaEeHH8Spqt4rayGhjXTce"}]}]}} {"custom_id": "Sylhet_33345_14364_33858_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XPcsuCcGrmXbRZ29S4Pywx"}]}]}} {"custom_id": "Narsingdi_14877_9234_15390_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5B4qGqfAfy4gRA7cxmRHhh"}]}]}} {"custom_id": "Manikganj_14877_12312_15390_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-D13VKwQ5WaNozU6wtY8yBD"}]}]}} {"custom_id": "Rangamati_17442_36936_17955_37449.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-15dqMdh4etw69uR9bZvBvy"}]}]}} {"custom_id": "Sylhet_4617_1539_5130_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6ZY33LFxszwiM9NAd7DeB5"}]}]}} {"custom_id": "Satkhira_6669_57456_7182_57969.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Lpxf8qL5ryiVz3ALFG32SQ"}]}]}} {"custom_id": "Dinajpur_6669_12312_7182_12825.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Csh8a7csnVBj2GDXHsDjXt"}]}]}} {"custom_id": "Lalmonirhat_13851_24111_14364_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MLkvJsHeo6EHnQ9F9jvAKo"}]}]}} {"custom_id": "Habiganj_19494_16416_20007_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VLDQFS1Hfdkid6AD1Z29ec"}]}]}} {"custom_id": "Tangail_16416_16929_16929_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XEumvEZMm8wwSWEGkCHMQQ"}]}]}} {"custom_id": "Rangpur_24111_9234_24624_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DSzmkzUj3fZ2focpp5haer"}]}]}} {"custom_id": "Comilla_17442_22059_17955_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3LHKMcNFzS6xoJaNdf26Pr"}]}]}} {"custom_id": "Bhola_5130_16416_5643_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4AqnatqhUkYsjsHfJWoqYk"}]}]}} {"custom_id": "Comilla_21546_20007_22059_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BKAAxfM4EXyZhgQ6C3UEky"}]}]}} {"custom_id": "Thakurgaon_14364_23598_14877_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XGy7DFU9FMKgEuvPJgqAfL"}]}]}} {"custom_id": "Narail_3078_13338_3591_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V7NRqG5Qi3p4KWudBtLYn4"}]}]}} {"custom_id": "Panchagarh_8208_5643_8721_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9NJm3D9i7tuVjE3Vt3BrDG"}]}]}} {"custom_id": "Munshiganj_17442_13851_17955_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GhgSBL66KPauRoF3HPE4n5"}]}]}} {"custom_id": "Netrakona_12825_513_13338_1026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-12GtTJuo9G3R2MrtMB2kTf"}]}]}} {"custom_id": "Jhenaidah_17442_7695_17955_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-AG7qMEgrmVfpQ7ecyFMa9h"}]}]}} {"custom_id": "Mymensingh_9234_33345_9747_33858.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RBX2dwaKMHuyxuAmp2wmSp"}]}]}} {"custom_id": "Kushtia_23085_15903_23598_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-So59u6Ga3ack4iYum61U5B"}]}]}} {"custom_id": "Barisal_13338_11286_13851_11799.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XNPCcQPWEtCNQLp1EhjWzg"}]}]}} {"custom_id": "Chittagong_22572_34884_23085_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4SvFfLvueNGfau9ZE4JBM5"}]}]}} {"custom_id": "Sunamganj_31806_27702_32319_28215.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X9YJNDnitUTT4JcNJ19q2B"}]}]}} {"custom_id": "Nawabganj_10773_8721_11286_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GtAoYcXgDLmMgqVr3zXBaP"}]}]}} {"custom_id": "Rangamati_9234_1026_9747_1539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LmyPcDT5BN9wKEgEXqE9QD"}]}]}} {"custom_id": "Khagrachhari_8208_29241_8721_29754.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Sm6HE2BTYA7EmtP6es8sA6"}]}]}} {"custom_id": "Pirojpur_4617_34371_5130_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5pWtcpq5zeMtAH44RkeC72"}]}]}} {"custom_id": "Narsingdi_6156_7695_6669_8208.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-7GFTiCgUEK73PZCHkWxesw"}]}]}} {"custom_id": "Lalmonirhat_14877_28215_15390_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L9Twai39deNVcXRygru9Km"}]}]}} {"custom_id": "Thakurgaon_10773_0_11286_513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-RJFBBrG1xTS2XmZqYFXB5j"}]}]}} {"custom_id": "Satkhira_21033_50274_21546_50787.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-V4McauXG2kjSu6aUCnVuGS"}]}]}} {"custom_id": "Chandpur_9747_12825_10260_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8ojeUBznmY9QyP8zG9HrQq"}]}]}} {"custom_id": "Manikganj_24624_10260_25137_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XqHciSai4eH6Mfa6xu6jte"}]}]}} {"custom_id": "Jhalokati_7695_1539_8208_2052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Nw1cwijhxKca3J8kiQmx6b"}]}]}} {"custom_id": "Bogra_32832_7182_33345_7695.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5ah9CjJ74MhPRESsixHUs5"}]}]}} {"custom_id": "Tangail_23598_25650_24111_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-W3TVxfsSBy8YTYT43ihpX8"}]}]}} {"custom_id": "Gazipur_16416_4617_16929_5130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BvfTXoJqWAVkf9yGeyXjX5"}]}]}} {"custom_id": "Pabna_19494_22572_20007_23085.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Duj3wnC8Vt4UskLnD849BD"}]}]}} {"custom_id": "Netrakona_11799_27189_12312_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-EGSuh14AN2h8EK6aXgRsfc"}]}]}} {"custom_id": "Sunamganj_17955_10260_18468_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4sFmdC28o8ZxaFzFQJaShK"}]}]}} {"custom_id": "Rangamati_19494_58482_20007_58995.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XEkNKB41ggF9h1eNuYHZfv"}]}]}} {"custom_id": "Sirajganj_17955_25650_18468_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VntH9nJynxqKPt9Zxd2EiN"}]}]}} {"custom_id": "Bogra_28215_15903_28728_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Bq6rPo8SFvwDovpBj9r4wW"}]}]}} {"custom_id": "Nilphamari_1026_13338_1539_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4hM1ydmZGzQEYKFGgZkdn8"}]}]}} {"custom_id": "Natore_18981_3078_19494_3591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GrYfMD8gU7XibMkwRzsBM5"}]}]}} {"custom_id": "Thakurgaon_4617_13338_5130_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X3yehD6xvw2VFLVjFAdnnf"}]}]}} {"custom_id": "Khagrachhari_9747_26163_10260_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KM4BnMLthzgABEFiRFr4ru"}]}]}} {"custom_id": "Bandarban_1026_26163_1539_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1BAJcUkaFsEDfSGUTUQSwK"}]}]}} {"custom_id": "Sirajganj_15390_34371_15903_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ETGabZsAAiLD3ja47MgKht"}]}]}} {"custom_id": "Naogaon_21546_20520_22059_21033.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-4xRBgFxWrm3rcwWmLhueDK"}]}]}} {"custom_id": "Sylhet_15903_15390_16416_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3Q9xFRrFQmtWWxf7HQoqKW"}]}]}} {"custom_id": "Khagrachhari_4617_40527_5130_41040.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WCyyLnmutHGguVM7ZtCovw"}]}]}} {"custom_id": "Bandarban_12312_17955_12825_18468.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LyApsXZ4FfBf3eV3iMXwgU"}]}]}} {"custom_id": "Rangamati_12825_26676_13338_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BGEgzdFBQ7hSQ743GbH88T"}]}]}} {"custom_id": "Rajshahi_27189_22059_27702_22572.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5DLQujuZhNtcHUCMGp9dXs"}]}]}} {"custom_id": "Sunamganj_35397_14364_35910_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Fq2JK6r7TaWvKAGYnumK8Q"}]}]}} {"custom_id": "Mymensingh_11286_38988_11799_39501.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FSQNDPf9EWoY2BbhXXWkLA"}]}]}} {"custom_id": "Bandarban_22059_34371_22572_34884.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-GKz88XjQCxoRcYsJU9BxAD"}]}]}} {"custom_id": "Sirajganj_15390_12825_15903_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DhfNvxQuPqYgVAxN5nE6nR"}]}]}} {"custom_id": "Faridpur_3078_6156_3591_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-C95SHjYwmKWZH8g9mPiuaN"}]}]}} {"custom_id": "Kurigram_4104_24624_4617_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Whe9DDFcmVRFjgLhcUCrXP"}]}]}} {"custom_id": "Gazipur_17955_10773_18468_11286.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-22Mq9BNmeciDE4JdH4cpkb"}]}]}} {"custom_id": "Pirojpur_7182_24624_7695_25137.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-35NGDHJWhL7ZNp4KqzWYiA"}]}]}} {"custom_id": "Noakhali_11799_15390_12312_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BNyqvNoEAJxgoqjhKzNafT"}]}]}} {"custom_id": "Sunamganj_14877_4104_15390_4617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QE7VfFdR6gHPFm14ij4SfW"}]}]}} {"custom_id": "Mymensingh_8208_11799_8721_12312.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-K9STcu6jt8Y4C8ZEN6nLfJ"}]}]}} {"custom_id": "Satkhira_20520_52839_21033_53352.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-88wn8aVRQ8AZDK9DACKbbE"}]}]}} {"custom_id": "Rangamati_13338_64125_13851_64638.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KYExgsdbT3AG9rKj6EKxr4"}]}]}} {"custom_id": "Khagrachhari_11799_17442_12312_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-DscWmyqTFMEf3Xaan7ACyu"}]}]}} {"custom_id": "Netrakona_14364_27189_14877_27702.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-QedR8674WzevLajMiqe4Ri"}]}]}} {"custom_id": "Patuakhali_7695_6156_8208_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2edUP5SDeHnbMjoBgBi8MK"}]}]}} {"custom_id": "Bandarban_9747_38988_10260_39501.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MaH8ToBAsRPnwhXcrSJa1R"}]}]}} {"custom_id": "Rangamati_29241_85671_29754_86184.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HjfyJ32TpDXG28bXVm47NY"}]}]}} {"custom_id": "Mymensingh_16929_18981_17442_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LGiBxuoG5nNnevM5pygFhA"}]}]}} {"custom_id": "Mymensingh_15390_24111_15903_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ET8SQGnhjV9ZkeuxJmGDzN"}]}]}} {"custom_id": "Satkhira_7695_23598_8208_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VnnjKSo2Ps1csF1w7fb3Py"}]}]}} {"custom_id": "Rangamati_15903_39501_16416_40014.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-VzJfaHrr4XkjURguFDa6eb"}]}]}} {"custom_id": "Nawabganj_9234_16929_9747_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-LjDEPZpDiMNutev6dKKfxk"}]}]}} {"custom_id": "Netrakona_24624_16416_25137_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Lsi4r75NNuei1BrECskVHw"}]}]}} {"custom_id": "Faridpur_4617_10260_5130_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-L7xTqh1qDyD6w3ocKgFjrA"}]}]}} {"custom_id": "Patuakhali_10773_16929_11286_17442.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1Mv2JEi5D4FHQuZLHwF4Gk"}]}]}} {"custom_id": "CoxsBazar_20520_53865_21033_54378.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-HQcGRPSqaNyp36gsmdknk9"}]}]}} {"custom_id": "Sylhet_9234_15903_9747_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JzJWusoKiVUP1x8H6DqEQ9"}]}]}} {"custom_id": "Brahamanbaria_22059_16416_22572_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9wXQRb2rFxhiaHiEeGPHMU"}]}]}} {"custom_id": "Madaripur_11799_19494_12312_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WHp5SBAynuWt8EE5SD3ggr"}]}]}} {"custom_id": "Bandarban_8208_18468_8721_18981.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Cpc2JGQnyzFRP25qC6YFPJ"}]}]}} {"custom_id": "Panchagarh_15903_14877_16416_15390.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6DtLSbUnrSjAzQRb8KpWwD"}]}]}} {"custom_id": "Rangamati_16929_61560_17442_62073.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JsBGB56Ryc6MX41LbxRDmk"}]}]}} {"custom_id": "Dinajpur_23085_34884_23598_35397.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-X5u9CboDPVZqR7uFHJUHTM"}]}]}} {"custom_id": "Netrakona_17442_5643_17955_6156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Dir6N3pg4AivjBTZaDELN3"}]}]}} {"custom_id": "Sherpur_14877_13851_15390_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1PFWNQJ7rEuzAuhkkwTqmo"}]}]}} {"custom_id": "Narail_7695_12825_8208_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-859PQU98xGPCEpKNowYmQ9"}]}]}} {"custom_id": "Jhenaidah_15903_13338_16416_13851.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Uz9r7dLuhm18iXUQCD1aCR"}]}]}} {"custom_id": "Gopalganj_5643_17442_6156_17955.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XG2z2M7MKhoNT6T7t4uMzC"}]}]}} {"custom_id": "Rangpur_17442_6156_17955_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MadYDRfNkzwGRS9mc8Jche"}]}]}} {"custom_id": "Sunamganj_18981_24111_19494_24624.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WBxzEYEG4NnfNRznL3DMCK"}]}]}} {"custom_id": "Chuadanga_10260_9234_10773_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-6oSnSNyq4WNBePr62xpUME"}]}]}} {"custom_id": "Bhola_4104_45144_4617_45657.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CEL6wHe8YpsvjkcvqJTBBT"}]}]}} {"custom_id": "Natore_6669_18981_7182_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5f7PwQ3XL6FhB9CRwStPu2"}]}]}} {"custom_id": "Chittagong_13851_6156_14364_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NAZTvyoJUAdxXugWdwJK3o"}]}]}} {"custom_id": "Noakhali_5130_13851_5643_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KW3nYQz6a1PFRNpHoNWM8f"}]}]}} {"custom_id": "Khulna_15390_26676_15903_27189.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5kL3zJs2cRhqxfxbvTAY23"}]}]}} {"custom_id": "Maulvibazar_27189_6156_27702_6669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Hgrea5kf4HBgHjZAaFpTKg"}]}]}} {"custom_id": "Netrakona_6156_15390_6669_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BygNZofm4p2AyRfc88363s"}]}]}} {"custom_id": "Kishoreganj_13851_13851_14364_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JzRL8SpUjZJ8YMmh6w4TPB"}]}]}} {"custom_id": "Manikganj_12312_9234_12825_9747.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MGt74ey2vqgqVxWa1CvQYm"}]}]}} {"custom_id": "Patuakhali_14364_19494_14877_20007.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-MCk7UvXLXaPUsKYL5562Xv"}]}]}} {"custom_id": "Maulvibazar_24624_15390_25137_15903.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-3iQjsiwYoc6y16AVehkQSW"}]}]}} {"custom_id": "Jessore_26676_18981_27189_19494.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-BdazxbHG7wDsWzpvAkD9m3"}]}]}} {"custom_id": "Khulna_19494_6669_20007_7182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-WwPF2dXgTFGZcCfMkt3sb9"}]}]}} {"custom_id": "Barguna_3591_25650_4104_26163.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KunCVizwhwkP7yyAnumGp7"}]}]}} {"custom_id": "CoxsBazar_15390_15903_15903_16416.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-9uuy87xyYxmgqsRhUyWJT8"}]}]}} {"custom_id": "Dinajpur_5643_12825_6156_13338.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-NbXprGY91bqoYcF5nQiF9J"}]}]}} {"custom_id": "Comilla_31806_30267_32319_30780.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-2K8dXfUA1Qv97ModPXZNB2"}]}]}} {"custom_id": "Panchagarh_15390_20007_15903_20520.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-1JFHMphn5e4GpfHLNMphef"}]}]}} {"custom_id": "Pirojpur_2052_9747_2565_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-M9XWmwxx6DEJCVhQaWhKwu"}]}]}} {"custom_id": "Chittagong_29241_44118_29754_44631.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-FHMKNKKCpkzmiicE8hukFS"}]}]}} {"custom_id": "Bandarban_17955_23598_18468_24111.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Ua2SVHQXYh2FFih4abc9im"}]}]}} {"custom_id": "Sirajganj_20007_36423_20520_36936.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-XmCQxDTPbj4iR9pgQ8Pjd7"}]}]}} {"custom_id": "Sirajganj_23085_28215_23598_28728.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-JjXcUDiRzuYZzGwCkRcmDd"}]}]}} {"custom_id": "Khagrachhari_10773_2052_11286_2565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-KPZifKXFD2rHrSvKcX5gkA"}]}]}} {"custom_id": "Natore_16416_8721_16929_9234.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-URP4Lzbsw9o8pKu3KzZSe2"}]}]}} {"custom_id": "Maulvibazar_22059_14364_22572_14877.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-ULaW2VYJvZXJwigW8WzCsP"}]}]}} {"custom_id": "Kishoreganj_26676_13851_27189_14364.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Gt2xH3fim5jRUB8fDjgV74"}]}]}} {"custom_id": "Habiganj_11799_10260_12312_10773.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-Nuyu5h4waRi6ABauzvdGAX"}]}]}} {"custom_id": "Gazipur_21546_16416_22059_16929.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-CPDYqEdJ69iy32uxGZmzQg"}]}]}} {"custom_id": "Khulna_14877_9747_15390_10260.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-5onZQ8RY5BjZQ5xrh7CYeZ"}]}]}} {"custom_id": "Netrakona_28728_26163_29241_26676.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"farmland\": [[10, 20, 50, 60]], \"built-up\": [[60, 10, 90, 40]], \"meadow\": [], \"forest\": [[5, 70, 30, 95]], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "image_url": "openai://file-8Zv231AhxGSP2Bjze5nsSc"}]}]}}