Current face clustering behavior is overly restrictive and often fails in real-world usage.
Observed behavior:
- Faces are only clustered when images are near-identical or exact duplicates
- Same person across different poses, expressions, angles, or lighting frequently remains unclustered
- Adding new folders does not reliably trigger clustering without manual intervention
Expected behavior:
- Same person should be clustered across reasonable pose, expression, angle, and lighting variations
- Face detection and clustering should run automatically on folder addition
Current face clustering behavior is overly restrictive and often fails in real-world usage.
Observed behavior:
Expected behavior: