Page MenuHomePhabricator

Identify faces in images on commons
Open, Needs TriagePublic

Description

An improvement on the categorisation system on commons. Can we execute face identification model (machine learning) to the images on commons and suggest tags / categories for images?

Present state
The only thing where machine learning is being used currently (significantly) is NSFW.

Proposed
There are huge no. of images that are left uncategorised or categorised but not categorised / tagged with the person's name or categorised with person's name but not all the notable ones (in case of images having multiple people).
Identifying them can help in finding free use images of people whose articles don't have one on various wikis and improve categorisation on commons. Plus this will reduce huge workload of those doing the maintenance work manually without any assistance. And if the generated encoding is made open-source, that would be the greatest support to tech community working on machine learning and face identification.

Github library (I used for my website)
https://github.com/anubhavshrimal/FaceRecognition

See also
https://meta.wikimedia.org/wiki/WikiAI

Event Timeline

Could you provide more specific use cases, please?
I'm asking as this topic might be controversial or not.

Looks like this was brought up in https://commons.wikimedia.org/wiki/Commons:Help_desk#Face_recognition originally.