Suggestion: Edit face recognition clusters manually

[Felix-Franz]

Is your feature request related to a problem? Please describe.

Sometimes a face is wrongly assigned to a cluster (e.g. person B is recognized as person A).

Describe the solution you'd like

Since a 100% correct recognition is not feasible, I propose a manual possibility to add or remove faces to persons.

Concrete examples:

• Person B, who was wrongly recognised as person A, should be manually assigned to person B (if person B does not exist, it should be created).
• Person B and person A are the same person, it should be possible to manually merge them.
• A person recognised and accepted as person A should be subsequently removed from the recognition cluster (e.g. because he was wrongly tagged).

Describe alternatives you've considered

To prevent similar faces from being assigned to the same person, it would be possible to increase the sensitivity of face recognition (clustered faces need to be closer to each other).

Additional context Maybe also include different pictures of the same person to the demo to demonstrate that face recognition works cross-images.

[deviantintegral]

This is a particularly important feature when indexing a library of family photos. Siblings, or photos of different people but at the same age can be easily classified as the same person. As is, there's no way to either remove the incorrect face on a photo, or to split it into "this is a new or different face".

[nickbreen]

I'm experiencing the opposite effect.

With a few 10k's of family photos and I have a huge list of "Unknown person 123456" which are all the same few people.

It would be useful to configure the clustering to be slightly more aggressive.