Home

winnow

winnow pulls photos from Immich, selects diverse high-quality subsets using AI embeddings, and delivers them as training data for Frigate. It supports face recognition training (aligned crops uploaded directly to Frigate's API) and object classification training (YOLO-detected crops saved to disk).

It runs fully headless in Docker, is configured entirely through environment variables, and can run on a schedule.

Pages

• Setup — installation, Docker Compose configuration, GPU passthrough, environment variables
• Troubleshooting — common failures and fixes
• FAQ — frequently asked questions

Quick Links

• GitHub Repository
• CHANGELOG
• compose.yml
• .env.example