ilia/punimtag
1
1
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Phase 1: fix face quality sharpness scoring #89

Merged
ilia merged 1 commits from fix/face-quality-sharpness-phase1 into master 2026-08-05 10:07:58 -05:00
Conversation 0 Commits 1 Files Changed 4
+200 -13
ilia commented 2026-08-05 10:04:05 -05:00
Owner
Copy Link
Copy Source

Summary

  • Fix calculate_face_quality_score so sharpness uses Laplacian variance of the face crop (was variance of the 3×3 kernel — always ~constant)
  • Add face_laplacian_variance helper + tests/test_face_quality_score.py
  • Doc: docs/FACE_MATCH_ERROR_TAGGING.md checklist for sampling ~100 Auto-Match/Identify misses
  • ROADMAP: face recognition accuracy phases 1–6

Note

Existing DB quality_score values are unchanged until faces are re-processed (Phase 3).

Test plan

  • pytest tests/test_face_quality_score.py -v --noconftest (4 passed)
  • ruff check backend/services/face_service.py
  • CI green
## Summary - Fix `calculate_face_quality_score` so sharpness uses Laplacian variance of the **face crop** (was variance of the 3×3 kernel — always ~constant) - Add `face_laplacian_variance` helper + `tests/test_face_quality_score.py` - Doc: `docs/FACE_MATCH_ERROR_TAGGING.md` checklist for sampling ~100 Auto-Match/Identify misses - ROADMAP: face recognition accuracy phases 1–6 ## Note Existing DB `quality_score` values are unchanged until faces are re-processed (Phase 3). ## Test plan - [x] `pytest tests/test_face_quality_score.py -v --noconftest` (4 passed) - [x] `ruff check backend/services/face_service.py` - [ ] CI green
ilia added 1 commit 2026-08-05 10:04:05 -05:00
Fix face quality sharpness: Laplacian on face crop (Phase 1).
CI / skip-ci-check (pull_request) Successful in 30s
Details
CI / python-lint (pull_request) Successful in 32s
Details
CI / docker-ci (pull_request) Successful in 33s
Details
CI / secret-scan (pull_request) Successful in 36s
Details
CI / viewer-unit (pull_request) Successful in 1m44s
Details
CI / e2e (pull_request) Successful in 1m54s
Details
CI / admin-unit (pull_request) Successful in 2m0s
Details
7856be47ff 
Sharpness previously used variance of the kernel constant, so every face
got ~0 sharpness weight. Compute Laplacian variance on the grayscale crop;
add unit tests and an error-tagging checklist for follow-on accuracy work.
ilia scheduled this pull request to auto merge when all checks succeed 2026-08-05 10:05:57 -05:00
ilia merged commit dfd0c582e6 into master 2026-08-05 10:07:58 -05:00
ilia deleted branch fix/face-quality-sharpness-phase1 2026-08-05 10:07:58 -05:00
Sign in to join this conversation.
Reviewers
No Reviewers
Labels
No items
No labels
Milestone
No items
No Milestone
Projects
Clear projects
No projects
Assignees
Clear assignees
ira (Ira) noam tanyar09 ilia (Ilia)
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: ilia/punimtag#89