Upload folder using huggingface_hub

README.md

@@ -1,214 +1,205 @@
----
-license: apache-2.0
-task_categories:
-- question-answering
-language:
-- zh
-size_categories:
-- n<1K
----
-# Titan CV Agent Benchmark
-
-[[中文版]](README_CN.md)
-
-**The Titan CV Benchmark is primarily designed to evaluate the performance of AGENTS in the field of computer vision (CV).** We have collected more than 200 test examples to comprehensively test the performance of the agent, particularly their capability to solve problems step by step.
-
-
-
-## Basic principles
-
-In order to distinguish the traditional benchmark of VLM, all collected samples will follow the following principles:
-
-① **Diverse modalities**, including various types of images and videos, and **rich application domains** such as industry, medicine, agriculture, environment, society, culture, sports, and scientific research.
-
-② Each problem will try to maintain its complexity, and **it cannot be solved in a single step** (e.g., using a VLM alone). Instead, it requires multi-step serial thinking.
-
-③ Each problem will focus on the intelligence of the intelligent agent. **The problems should appear simple to humans but be complex for machines.** Except in the medical field, it can be easily understood by any inexperienced adult, but it needs to fully mobilize its thinking and cognitive abilities. **For machines, traditional advantages such as memorization, calculation, or speed are not helpful in solving these problems**, and they need spatial cognition or advanced understanding abilities similar to humans.
-
-④ Each question is complex, but **the answer is simple and verifiable**, avoiding the ambiguity and complexity of open-ended question evaluation. If it is a question similar to the position difference, certain options will be given to avoid ambiguity, but at the same time, in order to prevent the intelligent agent from guessing the correct answer, the alternative answers are usually set to more than 6 to reduce the risk of guessing.
-
-
-
-## Quantity Classification
-
-This test set collects most of the CV problems on the market and divides them into **4 categories and 22 subcategories**:
-
-### Industrial Manufacturing and People's Livelihood Technology (35%)
-
-1. **Industrial Manufacturing** (10%) (Industrial Automation Detection / Packaging and Logistics Sorting System / Building Safety Monitoring / Energy and Power Inspection)
-
-2. **Public Safety** (8%) (Security Monitoring and Abnormal Behavior Detection / Face and License Plate Recognition / Crowd Density Estimation and Traffic Analysis / Fireworks and Fire Automatic Detection and Warning)
-
-3. **Medical Health** (8%) (Medical Image Analysis / Visual Assistance and Blind Navigation Technology / Personalized Rehabilitation)
-
-4. **Agricultural Production** (6%) (Pest and Disease Detection / Drone Image Processing / Plant Identification and Gardening Guidance / Crop Quality Inspection)
-
-5. **Environmental Monitoring** (3%) (Remote Sensing Image Analysis / Garbage Classification / Animal Identification and Ecological Protection / Soil and Water Quality Analysis / Disaster Emergency Response and Rescue)
-
-### Smart City and Smart Life (30%)
-
-6. **Fintech** (8%) (risk control / ticket identification and authenticity detection / blockchain / investment advice)
-7. **Transportation and logistics** (8%) (autonomous driving technology / intelligent traffic management / license plate recognition and parking management / logistics management / package management)
-8. **Smart terminals** (6%) (augmented reality and virtual reality / smart home / smart wearables / smart toys and game interaction / indoor navigation and mapping / human-computer interaction gesture and posture recognition)
-9. **E-commerce** (5%) (cross-border e-commerce / smart retail and automatic checkout / online shopping / human 3D modeling / virtual fitting)
-10. **Social life** (3%) (emotion recognition and user experience / content tagging / topic detection / intelligent advertising recommendation)
-
-### Cultural tourism and sports entertainment (20%)
-
-11. **Cultural education** (4%) (online classroom / educational marketing / smart education / art restoration / handwriting font recognition)
-12. **Travel guide** (4%) (smart park / scenic spot parking / life interconnection / Cultural and creative products)
-13. **Media and entertainment** (3%) (content creation / content review / video summary / intelligent editing / forgery detection / emoticon generation)
-14. **Sports technology** (3%) (human behavior analysis / sports analysis / sports tracking / somatosensory games)
-15. **Photo editing and beauty** (2%) (shooting optimization, beauty / smart album / makeup simulation / hair design)
-16. **Pet breeding** (2%) (pet monitoring / pet health care / smart feeding / pet interaction)
-17. **Art collection** (2%) (authenticity identification of cultural relics and jewelry / art value assessment)
-
-### Scientific research and professional fields (15%)
-
-18. **Scientific research** (4%) (marine life identification / animal migration research / psychological research)
-19. **Office workplace** (4%) (meeting records and automatic summaries / remote collaboration / intelligent attendance)
-20. **Government and judicial affairs** (3%) (government collaboration / smart human resources / evidence analysis / intelligent trial / Contract review)
-21. **Military and defense** (2%) (situational awareness / camouflage detection / enemy equipment identification)
-22. **Aerospace** (2%) (satellite image analysis / aircraft takeoff and landing monitoring / runway foreign object detection / pilot assistance system)
-
-**Note:** Data items in these categories will be collected gradually. The currently available public datasets only cover categories 1 through 8, and the remaining items will be gradually open sourced.
-
-
-
-## Quantity distribution
-
-|                                                        |                               | Relative Proportion (%) | Subtotal Proportion (%) | Absolute Quantity | Subtotal Quantity |
-| ------------------------------------------------------ | ----------------------------- | ----------------------- | ----------------------- | ----------------- | ----------------- |
-| **Industrial manufacturing and livelihood technology** | Industrial manufacturing      | 10                      | 35                      | 50                | 175               |
-|                                                        | Public security               | 8                       |                         | 40                |                   |
-|                                                        | Healthcare                    | 8                       |                         | 40                |                   |
-|                                                        | Agricultural production       | 6                       |                         | 30                |                   |
-|                                                        | Environmental monitoring      | 3                       |                         | 15                |                   |
-| **Smart city and smart life**                          | Fintech                       | 8                       | 30                      | 40                | 150               |
-|                                                        | Transportation and logistics  | 8                       |                         | 40                |                   |
-|                                                        | Smart terminals               | 6                       |                         | 30                |                   |
-|                                                        | E-commerce                    | 5                       |                         | 25                |                   |
-|                                                        | Social life                   | 3                       |                         | 15                |                   |
-| **Cultural tourism and sports entertainment**          | Culture and education         | 4                       | 20                      | 20                | 100               |
-|                                                        | Travel guide                  | 4                       |                         | 20                |                   |
-|                                                        | Media and entertainment       | 3                       |                         | 15                |                   |
-|                                                        | Sports technology             | 3                       |                         | 15                |                   |
-|                                                        | Photo editing and beauty      | 2                       |                         | 10                |                   |
-|                                                        | Pet breeding                  | 2                       |                         | 10                |                   |
-|                                                        | Art collection                | 2                       |                         | 10                |                   |
-| **Scientific research and professional fields**        | Scientific research           | 4                       | 15                      | 20                | 75                |
-|                                                        | Office                        | 4                       |                         | 20                |                   |
-|                                                        | Government and justice        | 3                       |                         | 15                |                   |
-|                                                        | Military and national defense | 2                       |                         | 10                |                   |
-|                                                        | Aerospace                     | 2                       |                         | 10                |                   |
-| Total                                                  | N/A                           | 100                     |                         | 500               |                   |
-
-**Note:** Data items in these categories will be collected gradually. Currently, the public dataset only involves 1 to 8 of them. The remaining items will be gradually open sourced.
-
-
-
-## Data format example
-
-### Data format
-
-Each data item contains 6 fields:
-
-```json
-"id": Data item number, named according to "categories_subcategories_items".
-"media_path": The path to the media file associated with the data item.
-"media_type": Media type that the data item depends on.
-"query": The question posed regarding the media content.
-"answer": Answer to the question.
-"note": Notes, such as a record of how the answer was obtained.
-```
-
-### Data example
-
-```json
-[
-    {
-        "id": "1_1_1",
-        "media_path": "media/01/1_1_1.mp4",
-        "media_type": "video",
-        "query": "有一个监控摄像头设置在生产线出料口，加工完的金属零件会沿着绿色倾斜平面滑落到下方接料盒，请统计一共有多少金属零件被生产出来？请回答一个整数，例如：10。请直接输出答案，不要输出任何符号、解释或说明。",
-        "answer": "11",
-        "note": "2+3+4+2=11"
-    },
-    {
-        "id": "1_1_2",
-        "media_path": "media/01/1_1_2.mp4",
-        "media_type": "video",
-        "query": "为了监督工人的产品专配过程，防止漏装、错装，工人的左手边、右手边和中间各有三堆零件，请阐述其零件拿取顺序，零件可以重复拿取。请按照：(中文单字,中文单字,中文单字)的格式直接回答，例如：(中,左,左)。请直接输出答案，不要输出任何符号、解释或说明。",
-        "answer": "(右,左,中)",
-        "note": "共计4个零件"
-    },
-    {
-        "id": "1_1_3",
-        "media_path": "media/01/1_1_3.mp4",
-        "media_type": "video",
-        "query": "请制作一个焊接检测系统，检测绿色电路板中零件数量和每个零件需要自动焊接多少次？请按照：(零件数量（整数）,每个零件需要自动焊接次数（整数）)的格式直接回答，例如：(10,1)。请直接输出答案，不要输出任何符号、解释或说明。",
-        "answer": "(60,2)",
-        "note": ""
-    }
-]
-```
-
-
-
-## Data Media Examples
-
-### Industrial Manufacturing
-
-![01_collage](imgs/01_collage.jpg)
-
-### Public safety
-
-![02_collage](imgs/02_collage.jpg)
-
-### Healthcare
-
-![03_collage](imgs/03_collage.jpg)
-
-### Agricultural production
-
-![04_collage](imgs/04_collage.jpg)
-
-### Environmental monitoring
-
-![05_collage](imgs/05_collage.jpg)
-
-### Financial technology
-
-![06_collage](imgs/06_collage.jpg)
-
-### Transportation and logistics
-
-![07_collage](imgs/07_collage.jpg)
-
-### Smart terminals
-
-![08_collage](imgs/08_collage.jpg)
-
-
-
-## Acknowledgments
-
-**This project is supported by:**
-
-***[Datacanvas - Empowering AI Native Businesses Worldwide](https://www.datacanvas.com/en/AboutUs)***
-
-> TEL: +86 400-805-7188
->
-> Email: [email protected]
->
-> Beijing Office: 6F&8F, Building C, HEYING Center, 10 Xiaoyingxi Road, Haidian District, Beijing, China
->
-> Shanghai Office: 10F, Building 1, SHIBEI TBC Center, Lane 1401, Jiangchang Road, Jing'an District, Shanghai, China
-
-
-
-## License
-
-**Titan CV Agent Benchmark uses the following license agreement:**
-
+# Titan CV Agent Benchmark
+
+[[中文版]](README_CN.md)
+
+**The Titan CV Benchmark is primarily designed to evaluate the performance of AGENTS in the field of computer vision (CV).** We have collected more than 200 test examples to comprehensively test the performance of the agent, particularly their capability to solve problems step by step.
+
+
+
+## Basic principles
+
+In order to distinguish the traditional benchmark of VLM, all collected samples will follow the following principles:
+
+① **Diverse modalities**, including various types of images and videos, and **rich application domains** such as industry, medicine, agriculture, environment, society, culture, sports, and scientific research.
+
+② Each problem will try to maintain its complexity, and **it cannot be solved in a single step** (e.g., using a VLM alone). Instead, it requires multi-step serial thinking.
+
+③ Each problem will focus on the intelligence of the intelligent agent. **The problems should appear simple to humans but be complex for machines.** Except in the medical field, it can be easily understood by any inexperienced adult, but it needs to fully mobilize its thinking and cognitive abilities. **For machines, traditional advantages such as memorization, calculation, or speed are not helpful in solving these problems**, and they need spatial cognition or advanced understanding abilities similar to humans.
+
+④ Each question is complex, but **the answer is simple and verifiable**, avoiding the ambiguity and complexity of open-ended question evaluation. If it is a question similar to the position difference, certain options will be given to avoid ambiguity, but at the same time, in order to prevent the intelligent agent from guessing the correct answer, the alternative answers are usually set to more than 6 to reduce the risk of guessing.
+
+
+
+## Quantity Classification
+
+This test set collects most of the CV problems on the market and divides them into **4 categories and 22 subcategories**:
+
+### Industrial Manufacturing and People's Livelihood Technology (35%)
+
+1. **Industrial Manufacturing** (10%) (Industrial Automation Detection / Packaging and Logistics Sorting System / Building Safety Monitoring / Energy and Power Inspection)
+
+2. **Public Safety** (8%) (Security Monitoring and Abnormal Behavior Detection / Face and License Plate Recognition / Crowd Density Estimation and Traffic Analysis / Fireworks and Fire Automatic Detection and Warning)
+
+3. **Medical Health** (8%) (Medical Image Analysis / Visual Assistance and Blind Navigation Technology / Personalized Rehabilitation)
+
+4. **Agricultural Production** (6%) (Pest and Disease Detection / Drone Image Processing / Plant Identification and Gardening Guidance / Crop Quality Inspection)
+
+5. **Environmental Monitoring** (3%) (Remote Sensing Image Analysis / Garbage Classification / Animal Identification and Ecological Protection / Soil and Water Quality Analysis / Disaster Emergency Response and Rescue)
+
+### Smart City and Smart Life (30%)
+
+6. **Fintech** (8%) (risk control / ticket identification and authenticity detection / blockchain / investment advice)
+7. **Transportation and logistics** (8%) (autonomous driving technology / intelligent traffic management / license plate recognition and parking management / logistics management / package management)
+8. **Smart terminals** (6%) (augmented reality and virtual reality / smart home / smart wearables / smart toys and game interaction / indoor navigation and mapping / human-computer interaction gesture and posture recognition)
+9. **E-commerce** (5%) (cross-border e-commerce / smart retail and automatic checkout / online shopping / human 3D modeling / virtual fitting)
+10. **Social life** (3%) (emotion recognition and user experience / content tagging / topic detection / intelligent advertising recommendation)
+
+### Cultural tourism and sports entertainment (20%)
+
+11. **Cultural education** (4%) (online classroom / educational marketing / smart education / art restoration / handwriting font recognition)
+12. **Travel guide** (4%) (smart park / scenic spot parking / life interconnection / Cultural and creative products)
+13. **Media and entertainment** (3%) (content creation / content review / video summary / intelligent editing / forgery detection / emoticon generation)
+14. **Sports technology** (3%) (human behavior analysis / sports analysis / sports tracking / somatosensory games)
+15. **Photo editing and beauty** (2%) (shooting optimization, beauty / smart album / makeup simulation / hair design)
+16. **Pet breeding** (2%) (pet monitoring / pet health care / smart feeding / pet interaction)
+17. **Art collection** (2%) (authenticity identification of cultural relics and jewelry / art value assessment)
+
+### Scientific research and professional fields (15%)
+
+18. **Scientific research** (4%) (marine life identification / animal migration research / psychological research)
+19. **Office workplace** (4%) (meeting records and automatic summaries / remote collaboration / intelligent attendance)
+20. **Government and judicial affairs** (3%) (government collaboration / smart human resources / evidence analysis / intelligent trial / Contract review)
+21. **Military and defense** (2%) (situational awareness / camouflage detection / enemy equipment identification)
+22. **Aerospace** (2%) (satellite image analysis / aircraft takeoff and landing monitoring / runway foreign object detection / pilot assistance system)
+
+**Note:** Data items in these categories will be collected gradually. The currently available public datasets only cover categories 1 through 8, and the remaining items will be gradually open sourced.
+
+
+
+## Quantity distribution
+
+|                                                        |                               | Relative Proportion (%) | Subtotal Proportion (%) | Absolute Quantity | Subtotal Quantity |
+| ------------------------------------------------------ | ----------------------------- | ----------------------- | ----------------------- | ----------------- | ----------------- |
+| **Industrial manufacturing and livelihood technology** | Industrial manufacturing      | 10                      | 35                      | 50                | 175               |
+|                                                        | Public security               | 8                       |                         | 40                |                   |
+|                                                        | Healthcare                    | 8                       |                         | 40                |                   |
+|                                                        | Agricultural production       | 6                       |                         | 30                |                   |
+|                                                        | Environmental monitoring      | 3                       |                         | 15                |                   |
+| **Smart city and smart life**                          | Fintech                       | 8                       | 30                      | 40                | 150               |
+|                                                        | Transportation and logistics  | 8                       |                         | 40                |                   |
+|                                                        | Smart terminals               | 6                       |                         | 30                |                   |
+|                                                        | E-commerce                    | 5                       |                         | 25                |                   |
+|                                                        | Social life                   | 3                       |                         | 15                |                   |
+| **Cultural tourism and sports entertainment**          | Culture and education         | 4                       | 20                      | 20                | 100               |
+|                                                        | Travel guide                  | 4                       |                         | 20                |                   |
+|                                                        | Media and entertainment       | 3                       |                         | 15                |                   |
+|                                                        | Sports technology             | 3                       |                         | 15                |                   |
+|                                                        | Photo editing and beauty      | 2                       |                         | 10                |                   |
+|                                                        | Pet breeding                  | 2                       |                         | 10                |                   |
+|                                                        | Art collection                | 2                       |                         | 10                |                   |
+| **Scientific research and professional fields**        | Scientific research           | 4                       | 15                      | 20                | 75                |
+|                                                        | Office                        | 4                       |                         | 20                |                   |
+|                                                        | Government and justice        | 3                       |                         | 15                |                   |
+|                                                        | Military and national defense | 2                       |                         | 10                |                   |
+|                                                        | Aerospace                     | 2                       |                         | 10                |                   |
+| Total                                                  | N/A                           | 100                     |                         | 500               |                   |
+
+**Note:** Data items in these categories will be collected gradually. Currently, the public dataset only involves 1 to 8 of them. The remaining items will be gradually open sourced.
+
+
+
+## Data format example
+
+### Data format
+
+Each data item contains 6 fields:
+
+```json
+"id": Data item number, named according to "categories_subcategories_items".
+"media_path": The path to the media file associated with the data item.
+"media_type": Media type that the data item depends on.
+"query": The question posed regarding the media content.
+"answer": Answer to the question.
+"note": Notes, such as a record of how the answer was obtained.
+```
+
+### Data example
+
+```json
+[
+    {
+        "id": "1_1_1",
+        "media_path": "media/01/1_1_1.mp4",
+        "media_type": "video",
+        "query": "有一个监控摄像头设置在生产线出料口，加工完的金属零件会沿着绿色倾斜平面滑落到下方接料盒，请统计一共有多少金属零件被生产出来？请回答一个整数，例如：10。请直接输出答案，不要输出任何符号、解释或说明。",
+        "answer": "11",
+        "note": "2+3+4+2=11"
+    },
+    {
+        "id": "1_1_2",
+        "media_path": "media/01/1_1_2.mp4",
+        "media_type": "video",
+        "query": "为了监督工人的产品专配过程，防止漏装、错装，工人的左手边、右手边和中间各有三堆零件，请阐述其零件拿取顺序，零件可以重复拿取。请按照：(中文单字,中文单字,中文单字)的格式直接回答，例如：(中,左,左)。请直接输出答案，不要输出任何符号、解释或说明。",
+        "answer": "(右,左,中)",
+        "note": "共计4个零件"
+    },
+    {
+        "id": "1_1_3",
+        "media_path": "media/01/1_1_3.mp4",
+        "media_type": "video",
+        "query": "请制作一个焊接检测系统，检测绿色电路板中零件数量和每个零件需要自动焊接多少次？请按照：(零件数量（整数）,每个零件需要自动焊接次数（整数）)的格式直接回答，例如：(10,1)。请直接输出答案，不要输出任何符号、解释或说明。",
+        "answer": "(60,2)",
+        "note": ""
+    }
+]
+```
+
+
+
+## Data Media Examples
+
+### Industrial Manufacturing
+
+![01_collage](imgs/01_collage.jpg)
+
+### Public safety
+
+![02_collage](imgs/02_collage.jpg)
+
+### Healthcare
+
+![03_collage](imgs/03_collage.jpg)
+
+### Agricultural production
+
+![04_collage](imgs/04_collage.jpg)
+
+### Environmental monitoring
+
+![05_collage](imgs/05_collage.jpg)
+
+### Financial technology
+
+![06_collage](imgs/06_collage.jpg)
+
+### Transportation and logistics
+
+![07_collage](imgs/07_collage.jpg)
+
+### Smart terminals
+
+![08_collage](imgs/08_collage.jpg)
+
+
+
+## Acknowledgments
+
+**This project is supported by:**
+
+***[Datacanvas - Empowering AI Native Businesses Worldwide](https://www.datacanvas.com/en/AboutUs)***
+
+> TEL: +86 400-805-7188
+>
+> Email: [email protected]
+>
+> Beijing Office: 6F&8F, Building C, HEYING Center, 10 Xiaoyingxi Road, Haidian District, Beijing, China
+>
+> Shanghai Office: 10F, Building 1, SHIBEI TBC Center, Lane 1401, Jiangchang Road, Jing'an District, Shanghai, China
+
+
+
+## License
+
+**Titan CV Agent Benchmark uses the following license agreement:**
+
 **Apache License Version 2.0**

README_CN.md

@@ -160,7 +160,7 @@ Titan 计算机视觉智能体评测基准主要用于测试CV领域的Agent性
 
 ### 环境监测
 
-![05_collage](D:\code\json_deal\imgs\05_collage.jpg)
+![05_collage](imgs/05_collage.jpg)
 
 ### 金融科技
 

titan_agent_benchmark_v3.json

@@ -91,7 +91,7 @@
         "id": "1_1_12",
         "media_path": "media/01/1_1_12.mp4",
         "media_type": "video",
-        "query": "为了建筑工地的工人安全，需要确保所有人都佩戴安全帽，请计算视频中多少人戴了安全帽，多少人没有戴安全帽？请按照：戴了安全帽的人数（整数）:没戴安全帽的人数（整数）的格式直接回答，例如：4:1。请直接输出答案，不要输出任何符号、解释或说明。",
+        "query": "为了建筑工地的工人安全，需要确保所有人都佩戴安全帽，请计算视频中多少人戴了安全帽，多少人没有戴安全帽？请按照：戴了安全帽的人数（整数）:没戴安全帽的人数（整数）的格式直接回答，例如：2:3。请直接输出答案，不要输出任何符号、解释或说明。",
         "answer": "5:1",
         "note": ""
     },
@@ -108,22 +108,22 @@
         "media_path": "media/01/1_1_14.mp4",
         "media_type": "video",
         "query": "视频中是自动元件组装流水线，需要计算设备在短时间安装了多少个电容？请回答一个整数，例如：1。请直接输出答案，不要输出任何符号、解释或说明。",
-        "answer": "5",
+        "answer": "7",
         "note": ""
     },
     {
         "id": "1_1_15",
         "media_path": "media/01/1_1_15.mp4",
         "media_type": "video",
-        "query": "为保证安全，要求机械手臂同时操作零件数不能大于2个，我们使用监控实时了解其操作情况，请问在视频所示的场景中，有几个零件在同时被操作？请回答一个整数，例如：1。请直接输出答案，不要输出任何符号、解释或说明。",
-        "answer": "3",
+        "query": "为保证安全，要求机械手臂同时操作零件数不能大于2个，我们使用监控实时了解其操作情况，请问在视频所示的场景中，有几个机械臂同时在操作？请回答一个整数，例如：1。请直接输出答案，不要输出任何符号、解释或说明。",
+        "answer": "4",
         "note": ""
     },
     {
         "id": "1_1_16",
         "media_path": "media/01/1_1_16.mp4",
         "media_type": "video",
-        "query": "汽车组装生产线中，需要将面板旋转一定角度，请判断旋转角度的范围。选项：A：0-45、B：46-90、C：91-135、D：136-180、E：181-225、F：226-270、G：271-315、H：316-360。表盘编号的后请直接输出答案，不要输出任何符号、解释或说明。",
+        "query": "汽车组装生产线中，需要将面板旋转一定角度，请判断旋转角度的范围。选项：A：0-45、B：46-90、C：91-135、D：136-180、E：181-225、F：226-270、G：271-315、H：316-360。请直接输出一个大写字母答案，例如：A，不要输出任何符号、解释或说明。",
         "answer": "E",
         "note": ""
     },
@@ -131,7 +131,7 @@
         "id": "1_1_17",
         "media_path": "media/01/1_1_17.mp4",
         "media_type": "video",
-        "query": "通过监控回答操作台经过哪个编号区域并最终停在了哪一个编号区域内？请按照：[经过的区域,...]:[停驻的区域,...]的格式直接回答，例如：[A01,A02]:[B01,B02,B03]。请直接输出答案，不要输出任何符号、解释或说明。",
+        "query": "通过监控回答操作台经过哪个编号区域并最终停在了哪一个编号区域内？请按照：[经过的区域,...]:[停驻的区域,...]的格式直接回答，例如：[A01,A02]:[B01]。请直接输出答案，不要输出任何符号、解释或说明。",
         "answer": "[G12]:[F12]",
         "note": ""
     },
@@ -171,7 +171,7 @@
         "id": "1_1_22",
         "media_path": "media/01/1_1_22.mp4",
         "media_type": "video",
-        "query": "视频中所示是两个手机屏幕的柔韧度测试，请用英文字母从左到右输出其弯折的程度像什么大写英文字母。回答格式：左边屏幕弯折像的字母;右边屏幕弯折像的字母，例如：A;B。请直接输出答案，不要输出任何符号、解释或说明。",
+        "query": "视频中所示是两个手机屏幕的柔韧度测试，请用英文字母表示从左到右的屏幕其弯折的程度像什么大写英文字母。回答格式：左边屏幕弯折像的字母;右边屏幕弯折像的字母，例如：A;B。请直接输出答案，不要输出任何符号、解释或说明。",
         "answer": "C;S",
         "note": ""
     },
@@ -179,7 +179,7 @@
         "id": "1_1_23",
         "media_path": "media/01/1_1_23.mp4",
         "media_type": "video",
-        "query": "有一个智能产线，在做硬度测试，其中硬度为手持表计所示，请输出指针靠近的最近的标示数值。请回答一个整数，例如：1。请直接输出答案，不要输出任何符号、解释或说明。",
+        "query": "有一个智能产线，在做硬度测试，其中硬度为手持表计所示，请输出指针最靠近的标示数值。请回答一个整数，例如：1。请直接输出答案，不要输出任何符号、解释或说明。",
         "answer": "30",
         "note": ""
     },
@@ -243,7 +243,7 @@
         "id": "1_1_31",
         "media_path": "media/01/1_1_31.jpg",
         "media_type": "image",
-        "query": "印刷厂在打印机口设置了摄像机，请告诉我货号是多少？是否有异物侵入？请按照：(货号（多位汉字、字母或数字）,是/否）的格式直接回答，例如：(ST2117,否)。请直接输出答案，不要输出任何符号、解释或说明。",
+        "query": "印刷厂在打印机口设置了摄像机，请告诉我货号是多少？是否有异物侵入？请按照：(货号（多位汉字、字母或数字）,是/否）的格式直接回答，例如：(AB1234,否)。请直接输出答案，不要输出任何符号、解释或说明。",
         "answer": "(ST2118,是)",
         "note": ""
     },
@@ -325,7 +325,7 @@
         "media_type": "image",
         "query": "杠铃片正在被打磨，我想知道这个杠铃片的重量，输出单位为公斤，保留1位小数，例如9.6。请直接输出答案，不要输出任何符号、解释或说明。",
         "answer": "20.4",
-        "note": ""
+        "note": "45磅 ≈ 20.4公斤"
     },
     {
         "id": "1_1_42",
@@ -355,8 +355,8 @@
         "id": "1_1_45",
         "media_path": "media/01/1_1_45.png",
         "media_type": "image",
-        "query": "观察图片，判断正在进行的操作。选项：A：装配、B：拆卸、C：清洁、D：检测、E：校准、F：电解。请回答一个大写字母，例如：D。请直接输出答案，不要输出任何符号、解释或说明。",
-        "answer": "A",
+        "query": "观察图片，推测最有可能的已完成组装的零件数量。请回答一个整数，例如：1。请直接输出答案，不要输出任何符号、解释或说明。",
+        "answer": "7",
         "note": ""
     },
     {
@@ -371,7 +371,7 @@
         "id": "1_1_47",
         "media_path": "media/01/1_1_47.png",
         "media_type": "image",
-        "query": "观察图片，判断该设备所在的环境。选项：Z：船舶制造车间、C：汽车制造车间、F：电子装配车间、W：食品加工车间。请回答一个大写字母，例如：C。请直接输出答案，不要输出任何符号、解释或说明。",
+        "query": "观察图片，判断该设备所在的环境。选项：Z：船舶制造车间、C：汽车制造车间、F：电子装配车间、W：食品加工车间、B：电池生产车间、D：制药生产车间。请回答一个大写字母，例如：C。请直接输出答案，不要输出任何符号、解释或说明。",
         "answer": "Z",
         "note": ""
     },
@@ -379,7 +379,7 @@
         "id": "1_1_48",
         "media_path": "media/01/1_1_48.png",
         "media_type": "image",
-        "query": "这是制造车灯中某个工业设备的局部特写图，请问图中夹具最有可能一次性夹取了多少个车灯元件？选项A：11个左右、选项B:2个左右、选项C：8个左右、选项D：5个左右、选项D：15个左右。请回答一个大写字母，例如：A。请直接输出答案，不要输出任何符号、解释或说明。",
+        "query": "图中是制造车灯中某个工业设备的局部特写图，请问图中夹具最有可能一次性夹取了多少个车灯元件？选项A：11个左右、选项B:2个左右、选项C：8个左右、选项D：5个左右、选项E：15个左右。请回答一个大写字母，例如：A。请直接输出答案，不要输出任何符号、解释或说明。",
         "answer": "C",
         "note": ""
     },
@@ -388,12 +388,12 @@
         "media_path": "media/01/1_1_49.png",
         "media_type": "image",
         "query": "画面中央黄色起重设备吊起的物体是什么？选项如下：A：塑料桶、B：盛钢桶、C：木桶、D：竹筒、E：箱子、F：金属条。请回答一个大写字母，例如：A。请直接输出答案，不要输出任何符号、解释或说明。",
-        "answer": "盛钢桶",
+        "answer": "B",
         "note": ""
     },
     {
         "id": "1_1_50",
-        "media_path": "media/01/1_1_50.png",
+        "media_path": "media/01/1_1_50.jpg",
         "media_type": "image",
         "query": "图中正在被操作的元件是什么？选项A：CPU、选项B：显卡、选项C：内存条、选项D：硬盘、选项E：主板。请回答一个大写字母，例如：A。请直接输出答案，不要输出任何符号、解释或说明。",
         "answer": "D",
@@ -419,7 +419,7 @@
         "id": "1_2_3",
         "media_path": "media/02/1_2_3.mp4",
         "media_type": "video",
-        "query": "请识别电梯内乘客的危险行为，请从以下选项中选择全部的危险行为。选项：A：扒电梯门、B：跳跃、C：徘徊、D：蹲下、F：打架、G：淹水、H：推电动车。请回答一位或多位大写字母序列，例如：ABC。请直接输出答案，不要输出任何符号、解释或说明。",
+        "query": "请识别电梯内乘客的危险行为，请从以下选项中选择全部的危险行为。选项：A：扒电梯门、B：跳跃、D：蹲下、F：打架、G：淹水、H：推电动车。请回答一位或多位大写字母序列，例如：ABC。请直接输出答案，不要输出任何符号、解释或说明。",
         "answer": "AD",
         "note": ""
     },
@@ -441,8 +441,8 @@
     },
     {
         "id": "1_2_6",
-        "media_path": "media/02/1_2_6.jpg",
-        "media_type": "image",
+        "media_path": "media/02/1_2_6.mp4",
+        "media_type": "video",
         "query": "我需要一个通用火灾检测系统，请识别视频中的可疑着火物，并判断其是（True）否（False）冒烟和着火。请按照：(着火物（多位汉字、字母或数字）,True/False,True/False)，表示可疑物是小汽车，有冒烟但没有着火的格式直接回答，例如：(汽车,False,True)。请直接输出答案，不要输出任何符号、解释或说明。",
         "answer": "(电动车,True,True)",
         "note": ""
@@ -460,7 +460,7 @@
         "media_path": "media/02/1_2_8.mp4",
         "media_type": "video",
         "query": "你是银行安全卫士，请识别办事窗口（视频红圈中）一共有几人，坐着人数和站立人数以及是否检测到什么可疑物体（选项：A：枪、B：刀、C：棍、D：无）。请按照：（总人数（整数）,站着的人（整数）,坐着的人（整数）,可疑物体（多位汉字、字母或数字））的格式直接回答，例如：(3,2,1,无)。请直接输出答案，不要输出任何符号、解释或说明。",
-        "answer": "(2,1,1,刀)",
+        "answer": "(3,1,2,刀)",
         "note": ""
     },
     {
@@ -545,7 +545,7 @@
     },
     {
         "id": "1_2_19",
-        "media_path": "media/02/1_2_19.jpg",
+        "media_path": "media/02/1_2_19.png",
         "media_type": "image",
         "query": "未系安全带会给司乘人员带来极大的安全隐患，请识别所有未系安全带的情况，从左到右检查，如果车辆的人员没有系安全带且车辆车牌可识别的情况下，请输出他的车牌号。回答格式：多位汉字、字母或数字,...，例如：你好ABC123,上海CBA321。请直接输出答案，不要输出任何符号、解释或说明。",
         "answer": "湘MD69262,鲁LQ611Q",
@@ -844,7 +844,7 @@
         "media_path": "media/04/1_4_6.mp4",
         "media_type": "video",
         "query": "我们使用固定机位拍摄农民摘果子的场景用以统计果实的摘取数量，请告诉我一共摘取了多少果子。请回答一个整数，例如：1。请直接输出答案，不要输出任何符号、解释或说明。",
-        "answer": "4",
+        "answer": "3",
         "note": ""
     },
     {
@@ -948,7 +948,7 @@
         "media_path": "media/05/1_5_1.mp4",
         "media_type": "video",
         "query": "为了研究海边垃圾种类和动物活动痕迹，请告诉我视频中有多少形态完好的棕色瓶子？请回答一个整数，例如：1。请直接输出答案，不要输出任何符号、解释或说明。",
-        "answer": "5",
+        "answer": "6",
         "note": ""
     },
     {
@@ -969,8 +969,8 @@
     },
     {
         "id": "1_5_4",
-        "media_path": "media/05/1_5_4.mp4",
-        "media_type": "video",
+        "media_path": "media/05/1_5_4.png",
+        "media_type": "image",
         "query": "请问我有一些废旧电池和旧书籍，应该分别放在什么颜色的垃圾箱内？请回答一个或多个中文单字，例如：黄橙。请直接输出答案，不要输出任何符号、解释或说明。",
         "answer": "红蓝",
         "note": ""
@@ -1011,7 +1011,7 @@
         "id": "1_5_9",
         "media_path": "media/05/1_5_9.jpg",
         "media_type": "image",
-        "query": "如图所示设立在某城市的三个环境监测站分别在：A：工地、B：高架路桥、C：市中心马路边，请告诉我PM10从低到高依次是？请回答一位或多位大写字母序列，例如：ABC。请直接输出答案，不要输出任何符号、解释或说明。",
+        "query": "如图所示设立在某城市的三个环境监测站可能在：A：工地、B：高架路桥、C：市中心马路边，请告诉我PM10从低到高依次是？请回答一位或多位大写字母序列，例如：ABC。请直接输出答案，不要输出任何符号、解释或说明。",
         "answer": "CBA",
         "note": ""
     },
@@ -1075,7 +1075,7 @@
         "id": "1_6_7",
         "media_path": "media/06/1_6_7.jpg",
         "media_type": "image",
-        "query": "根据图片回答问题，从北京到吉林的Z117次列车下铺比中铺贵多少钱？请回答一个整数，例如：10。请直接输出答案，不要输出任何符��、解释或说明。",
+        "query": "根据图片回答问题，从北京到吉林的Z117次列车下铺比中铺贵多少钱？请回答一个整数，例如：1。请直接输出答案，不要输出任何符号、解释或说明。",
         "answer": "9",
         "note": ""
     },
@@ -1107,8 +1107,8 @@
         "id": "1_6_11",
         "media_path": "media/06/1_6_11.jpg",
         "media_type": "image",
-        "query": "图中有两张百元大钞，已知一真一假，请你判断出假的百元大钞是哪一张？请回答上面那张或者下面那张。请直接输出答案，不要输出任何符号、解释或说明。",
-        "answer": "下面那张",
+        "query": "图中有两张百元大钞，已知一真一假，请你输出假的百元大钞的编号？请回答多位汉字、字母或数字，例如：你好ABC123。请直接输出答案，不要输出任何符号、解释或说明。",
+        "answer": "K21R782662",
         "note": ""
     },
     {
@@ -1123,7 +1123,7 @@
         "id": "1_6_13",
         "media_path": "media/06/1_6_13.mp4",
         "media_type": "video",
-        "query": "统计出视频中共进行了多少次的移动支付？输出格式为一个整数，例如0。请直接输出答案，不要输出任何符号、解释或说明。",
+        "query": "统计出视频中共进行了多少次的移动支付？输出格式为一个整数，例如1。请直接输出答案，不要输出任何符号、解释或说明。",
         "answer": "10",
         "note": ""
     },
@@ -1544,8 +1544,8 @@
         "note": ""
     },
     {
-        "id": "media/08/1_8_19",
-        "media_path": "1_8_19.jpg",
+        "id": "1_8_19",
+        "media_path": "media/08/1_8_19.jpg",
         "media_type": "image",
         "query": "统计出会议中除了主持人以外未关闭麦克风的有多少人？是否有人在进行屏幕共享？输出格式为：（未关闭麦克风人数，是否有人进行屏幕共享），例如：（2，是）或者（1，否）。请直接输出答案，不要输出任何符号、解释或说明。",
         "answer": "（3，否）",