Show HN: PicPick – AI-powered photo curator using CLIP and face recognition
5 days ago
- #photo-management
- #AI-clustering
- #open-source
- PicPick helps filter thousands of photos down to the best 200-300 for albums using AI.
- Key features include smart clustering, face recognition, timeline view, quick starring, shareable filters, and easy export.
- Smart clustering groups visually similar photos using CLIP embeddings.
- Face recognition automatically identifies people across all photos.
- Timeline view organizes photos by date and time with visual separators for events.
- Quick starring allows rapid photo selection with keyboard shortcuts.
- Shareable filters enable URL-based filtering for easy sharing.
- Easy export copies starred photos to a folder ready for albums.
- Workflow: Index photos → Cluster similar photos → Review → Star favorites → Export.
- Technical requirements: Python 3.11+, ~8GB RAM, photos in a folder.
- Installation involves cloning the repository, setting up a virtual environment, and installing dependencies.
- First run takes ~30-60 minutes for 5000 photos; subsequent runs are faster.
- Web UI allows navigation, starring, and exporting with keyboard shortcuts.
- Export options include copying, moving, and organizing by date folders.
- Customizable clustering parameters: DBSCAN_EPS, DBSCAN_MIN_SAMPLES, MIN_FACE_SIZE.
- Technologies used: Python, FastAPI, SQLite, CLIP, face_recognition, scikit-learn, Vanilla JS.
- Open-source contributions welcome for features like drag-and-drop reordering, cloud storage integration, and mobile-responsive UI.
- MIT License allows use for personal and commercial projects.