Facial Grouping ML verification

[duncaninnes]

Hi,

I can't seem to find an answer to this anywhere, but is there a way to verify the ML for a facial recognition?

e.g. 100 images are tagged with a specific face match. Those tags are best guess from the ML at this point. But could a user the strengthen the ML algorithms by verifying the face match? Removing faces is clearly done, but I don't see any way of elevating the scoring of a match based on user input.

User could be offered a selection when hovering over a the match image at the right of the page:

Confirm User
N/A
Not User

i.e. "this is definitely user X" rather than "ML thinks this is user X"

Don't know much about how the ML works, but would this be able to improve recognition by reinforcing positive matches?

Depends on user input of course. So no user input would be no different to the current state. But a user willing to go in and tick a box offered when hovering over

[ty-han]

This is absolutely necessary, as the current process is super tedious having to click into each image, correct faces, etc... It makes the facial recognition feature way too cumbersome to bother investing in.

[ty-han]

This should also be integrated with this request, which takes the user input and then has the engine refine the results based on the new user inputs, so that the entire fixing/refinement process is not manual.