Unable to reject face annotation in viewer

[happysnapper]

Anyone else having this problem?

I double click thumbnail image which has face annotations and it opens in viewer with the green squares with confirmed names in green and unconfirmed names in brown as expected. So far so good.
If I want to reject a name I click inside the green square and type R but instead of the name below being replaced with a ? as it should a random name appears and if I keep typing R the name keeps changing. Eventually the names change to a ? but it might require six, a dozen or more keystrokes. If I want to ignore the face and type I that works fine. This was never a problem with v.2020.
I have rebooted, and run diagnostics to no avail.
The link persons to files feature is a brilliant addition Mario.

[Mario]

name below being replaced with a ? as it should a random name appears and if I keep typing R the name keeps changing.

This is the expected behavior.

"Reject" means: "This is not this person".
When you reject a person, IMatch automatically searches for the next best matching person.
Usually this works, unless you have many persons with very similar faces and the training data is not good.

Ignoring means "Don't do anything with this face".

Maybe you should not press Reject multiple times but instead assign the correct person immediately?

[happysnapper]

I understand that logic Mario, this is what I thought was happening but as most suggestions  were so "off the wall" having little or no similar features it seemed wrong. Is there a reason this worked differently in ver.2020?   

[Mario]

Not really. IMatch 2021 just actually does try to find a matching person, 2020 did not - which was a bug.
I have tested this on a range of databases and image sets (from personal images to official face databases) and it works very well.

I suggest you make a quick check on the faces you have trained for the persons involved, to see if the training data is good enough.
Don't train too many faces. Enter the age for persons. Train a small number of faces for each age. 5 - 10 trained faces is usually more than sufficient.

Keep in mind: Face AI is maybe 20 years old, and real progress was made in the past 5 years. Compare that to the 5,000+ years (Bible etc.) or 100,000 years of face recognition tech optimization our brains use.
AI will not beat our human brains when it comes to face recognition any time soon.

[happysnapper]

Haha, I understand what you are saying and have to say generally I'm very impressed with face AI.
I do a lot of live music photography so have many thousands of different subjects. With ver.2020 I might know the name applied to a face was wrong but might not know, or  remember, the correct name at that moment in time so it was easier to reject the name. I wouldn't ignore because hopefully the AI would match with similar image which I later correctly named. I think what I will now need to do is ignore faces I know are wrong until a later time when I have researched the name and can manually enter it.
It was a bit mis concerting when Spike Edney (keyboard player with Queen) was named as Jamie Moses (guitarist with Paul Young's band) I could forgive the AI for that but the next AI choice was the lovely Madeline Bell of Blue Mink fame.  lol

IM. just gets better and better and yes I do sing your praises to anyone who will listen. :-)