# 1. Create conda environment
conda create -n bench python=3.12
conda activate bench

# 2. Install dependencies
pip install -r requirements.txt

# 3. Launch the application
python app.py

• Access URL: http://localhost:5001

Place JSON files containing "qacandidate" in the project's data/ folder.

Set the video download directory to the root static/videos/ folder. The expected format is static/videos/<video_name_folder>/*.mp4.

• Enter http://127.0.0.1:5001 to access the QA screening interface
• Load JSON: Select the JSON file to open from the top-left corner (the tool auto-detects files in the data folder; all changes are automatically synced to the loaded JSON file)
• Select segment: After loading a JSON, the segment list appears on the left. Click a segment to view its QA candidates
• Select QA: The QA list is in the center. Click the "Select" button to select the current QA
• Watch clips: The video player is on the right. After selecting a QA, you can play:
  • Full clip: Start time → End time
  • First half: Start time → Split point
  • Second half: Split point → End time
• Edit QA info: Following the QA screening workflow and guidelines above, you can edit QA details, set split points, delete QAs, or mark availability (available QAs are shown in green)

For each round of data collection, submit only the annotated JSON file, which stores all annotation information directly.

• Python 3.12
• FFmpeg
• Sufficient storage space for video downloads