{"id":28653866,"url":"https://github.com/tiger-ai-lab/videoeval-pro","last_synced_at":"2025-06-13T07:07:54.417Z","repository":{"id":294367753,"uuid":"984363742","full_name":"TIGER-AI-Lab/VideoEval-Pro","owner":"TIGER-AI-Lab","description":"More reliable Video Understanding Evaluation","archived":false,"fork":false,"pushed_at":"2025-06-10T14:07:10.000Z","size":6922,"stargazers_count":6,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-06-10T15:31:00.416Z","etag":null,"topics":["evaluation","multimodal","understanding","video"],"latest_commit_sha":null,"homepage":"https://tiger-ai-lab.github.io/VideoEval-Pro","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/TIGER-AI-Lab.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2025-05-15T20:09:40.000Z","updated_at":"2025-05-29T04:34:02.000Z","dependencies_parsed_at":"2025-05-20T05:19:37.282Z","dependency_job_id":"51a038cc-354a-4151-ad8a-a7078d546217","html_url":"https://github.com/TIGER-AI-Lab/VideoEval-Pro","commit_stats":null,"previous_names":["tiger-ai-lab/videoeval-pro"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/TIGER-AI-Lab/VideoEval-Pro","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/TIGER-AI-Lab%2FVideoEval-Pro","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/TIGER-AI-Lab%2FVideoEval-Pro/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/TIGER-AI-Lab%2FVideoEval-Pro/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/TIGER-AI-Lab%2FVideoEval-Pro/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/TIGER-AI-Lab","download_url":"https://codeload.github.com/TIGER-AI-Lab/VideoEval-Pro/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/TIGER-AI-Lab%2FVideoEval-Pro/sbom","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":259599331,"owners_count":22882357,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["evaluation","multimodal","understanding","video"],"created_at":"2025-06-13T07:07:53.815Z","updated_at":"2025-06-13T07:07:54.410Z","avatar_url":"https://github.com/TIGER-AI-Lab.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# VideoEval-Pro\nThis repository contains the evaluation code for the VideoEval-Pro. \nThe data is available on HuggingFace: [VideoEval-Pro](https://huggingface.co/datasets/TIGER-Lab/VideoEval-Pro)\n\n## Dataset Introduction\nVideoEval-Pro is a robust and realistic long video understanding benchmark containing open-ended, short-answer QA problems. The dataset is constructed by reformatting questions from four existing long video understanding MCQ benchmarks: Video-MME, MLVU, LVBench, and LongVideoBench into free-form questions.\n\n\n\n\nEach example in the dataset contains:\n- `video`: Name (path) of the video file\n- `question`: The question about the video content\n- `options`: Original options from the source benchmark\n- `answer`: The correct MCQ answer\n- `answer_text`: The correct free-form answer\n- `meta`: Additional metadata from the source benchmark\n- `source`: Source benchmark\n- `qa_subtype`: Question task subtype\n- `qa_type`: Question task type\n\n## Evaluation Steps\n\n1. **Download and Prepare Videos**\n   ```bash\n   # Download the dataset from HuggingFace\n   git lfs install\n   git clone https://huggingface.co/datasets/TIGER-Lab/VideoEval-Pro\n\n   # Navigate to videos directory\n   cd VideoEval-Pro/videos\n   \n   # Merge all split tar.gz files into a single archive\n   cat videos_part_*.tar.gz \u003e videos_merged.tar.gz\n   \n   # Extract the merged archive\n   tar -xzf videos_merged.tar.gz\n   \n   # [Optional] Clean up the split files and merged archive\n   rm videos_part_*.tar.gz videos_merged.tar.gz\n   \n   # After extraction, you will get a directory containing all videos\n   # The path to this directory will be used as --video_root in evaluation\n   # For example: 'VideoEval-Pro/videos'\n   ```\n\n2. **[Optional] Pre-extract Frames**\n   To improve efficiency, you can pre-extract frames from videos. The extracted frames should be organized as follows:\n   ```\n   frames_root/\n   ├── video_name_1/              # Video name\n   │   ├── 000001.jpg             # Frame images\n   │   ├── 000002.jpg\n   │   └── ...\n   ├── video_name_2/\n   │   ├── 000001.jpg\n   │   ├── 000002.jpg\n   │   └── ...\n   └── ...\n   ```\n\n   After frame extraction, the path to the frames will be used as `--frames_root`. Set `--using_frames True` when running the evaluation script.\n\n3. **Setup Evaluation Environment**\n   ```bash\n   # Clone the repository from the GitHub repository\n   git clone https://github.com/TIGER-AI-Lab/VideoEval-Pro\n   cd VideoEval-Pro\n   \n   # Create conda environment from requirements.txt (there are different env files for different models)\n   conda create -n videoevalpro --file *.yaml\n   conda activate videoevalpro\n   ```\n\n4. **Run Evaluation**\n   ```bash\n   cd VideoEval-Pro\n   \n   # Set PYTHONPATH\n   export PYTHONPATH=.\n   \n   # Run evaluation script with the following parameters:\n   # --video_root: Path to video files folder\n   # --frames_root: Path to video frames folder [For using_frames]\n   # --output_path: Path to save output results\n   # --using_frames: Whether to use pre-extracted frames\n   # --model_path: Path to model\n   # --device: Device to run inference on\n   # --num_frames: Number of frames to sample from video\n   # --max_retries: Maximum number of retries for failed inference\n   # --num_threads: Number of threads for parallel processing\n   \n   python tools/*_chat.py \\\n       --video_root \u003cpath_to_videos\u003e \\\n       --frames_root \u003cpath_to_frames\u003e \\\n       --output_path \u003cpath_to_save_results\u003e \\\n       --using_frames \u003cTrue/False\u003e \\\n       --model_path \u003cmodel_name_or_path\u003e \\\n       --device \u003cdevice\u003e \\\n       --num_frames \u003cnumber_of_frames\u003e \\\n       --max_retries \u003cmax_retries\u003e \\\n       --num_threads \u003cnum_threads\u003e\n\n   E.g.:\n   python tools/qwen_chat.py \\\n       --video_root ./videos \\\n       --frames_root ./frames \\\n       --output_path ./results/qwen_results.jsonl \\\n       --using_frames False \\\n       --model_path Qwen/Qwen2-VL-7B-Instruct \\\n       --device cuda \\\n       --num_frames 32 \\\n       --max_retries 10 \\\n       --num_threads 1\n   ```\n\n5. **Judge the results**\n   ```bash\n   cd VideoEval-Pro\n   \n   # Set PYTHONPATH\n   export PYTHONPATH=.\n   \n   # Run judge script *gpt4o_judge.py* with the following parameters:\n   # --input_path: Path to save output results\n   # --output_path: Path to judged results\n   # --model_name: Version of the judge model\n   # --num_threads: Number of threads for parallel processing\n   \n   python tools/gpt4o_judge.py \\\n       --input_path \u003cpath_to_saved_results\u003e \\\n       --output_path \u003cpath_to_judged_results\u003e \\\n       --model_name \u003cmodel_version\u003e \\\n       --num_threads \u003cnum_threads\u003e\n\n   E.g.:\n   python tools/gpt4o_judge.py \\\n       --input_path ./results/qwen_results.jsonl \\\n       --output_path ./results/qwen_results_judged.jsonl \\\n       --model_name gpt-4o-2024-08-06 \\\n       --num_threads 1\n   ```\n   **Note: the released results are judged by  *gpt-4o-2024-08-06***","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ftiger-ai-lab%2Fvideoeval-pro","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Ftiger-ai-lab%2Fvideoeval-pro","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ftiger-ai-lab%2Fvideoeval-pro/lists"}