Idea for Face Tagging

[Darthagnon]

Is your feature request related to a problem? Please describe.
Sorting by Faces is an awesome (and complicated) feature, usually done via machine learning algorithms in PhotoView, Photoprism, etc. However, there is an easier way...

Describe the solution you'd like
Google Picasa uses machine learning to recognise faces in photos, but the actual face tagging is done via either EXIF metadata with the person's name (if I recall correctly), or via multiple .picasa.ini files that list hashes matching to position on the photo (?) and a name, for example, a folder full of pictures of my face:

[Contacts2]
caefa70bc4c5dad2=Darren;;
[Profile Picture 3.jpg]
faces=rect64(1ee22a22b94da514),caefa70bc4c5dad2
backuphash=16963
[Profile Picture 3 medium.jpg]
faces=rect64(1e830830b94dc1dc),caefa70bc4c5dad2
backuphash=16963
[Profile Picture 1.JPG]
faces=rect64(34000aabbb99e3bb),caefa70bc4c5dad2
backuphash=1120
[Profile Picture 2.JPG]
faces=rect64(6000000de66ffff),caefa70bc4c5dad2
backuphash=37025
[admin.jpg]
faces=rect64(1e1407aeb999c1eb),caefa70bc4c5dad2
backuphash=64588

A "quick" way to add face recognition to Photofield, without implementing actual recognition models and related complications, would be to make it check for .picasa.ini and name metadata already present and map a People view from those. I don't know how useful this would be in the long term, but it was a silly idea I had of outsourcing the facial recognition to already-existing tools and just using what they generate. e.g. I believe Digikam uses name EXIF metadata by default (what Picasa calls Tools >> Options >> Name tags >> Store name tags in photo)

[SmilyOrg]

Good idea, thanks!

I was also considering CompreFace as it seems like a pretty complete solution, but I haven't tried it out yet. 😊