BUG: Face Clustering Quality Issues

[SiddharthJiyani]

Is there an existing issue for this?

• I have searched the existing issues

Description

The current face clustering implementation incorrectly groups faces, causing two main problems:

1. Multiple different people merged into single cluster - Photos of entirely different individuals are grouped together
2. Same person split across multiple clusters - When a person has frontal and profile (side-view) photos, they appear in separate clusters instead of one

This makes the face clustering feature unreliable and requires extensive manual correction.

How to Reproduce

1. Upload photos with multiple people and varying face angles.
2. Trigger clustering in Settings.
3. Observe incorrect groupings and splits

Benefits of Fixing

1. Accurate photo organization: Each person automatically gets their own cluster
2. Better search experience: Find all photos of a person in one place

• I will work on this issue.

Record

• I agree to follow this project's Code of Conduct

Checklist before Submitting

• Have you updated docs for it?
• Have you added unit tests?
• Have you made sure unit tests pass?
• Have you made sure code formatting is correct?
• Do Your changes passes all tests?

[coderabbitai]

📝 CodeRabbit Plan Mode

Generate an implementation plan and prompts that you can use with your favorite coding agent.

• Create Plan

Examples

• Example 1
• Example 2

_{You can configure auto-planning by selecting labels in the issue_enrichment configuration.}

🔗 Related PRs

#560 - feat: add global face reclustering option in settings with backend API and UI support [merged]

👤 Suggested Assignees

• Hemil36

---

🧪 Issue enrichment is currently in open beta.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord or schedule a call!

[SiddharthJiyani]

@rahulharpal1603 I want to work on it, can you please assign this issue to me?
Thanks.

[SiddharthJiyani]

@rahulharpal1603 I have raised a PR, please review it, the results for face clustering has improved a lot. Let me know if you see any problems, i tested rigorously from my side.
Thanks!