Face Recognition suggestions

[CVKen3]

I've been doing a lot of work with the face recognition system. Overall, it's really good. A few suggestions to maybe make the workflow more efficient?

1. If I add a manual face annotation (draw a box), it either comes up with a name or a question mark. Suggestion: If it doesn't generate a name, automatically pop up the Person Selector so that I can just start typing a name. Bonus points if the Person Selector chooses the first match if I type and hit ENTER! (I know, CTRL-1 does the trick, too.) 

2. The F6 key currently brings up a selector to either add a new face or rerun recognition on the image. Can we change it to just rerun facial recognition without asking? Currently you have to press F6, down arrow, enter key -- that will regenerate the faces. To just add a face, I use the Insert key.

The rest of this is just being picky, so not that important.

(PS: A little goofy that the close box on the Person Selector is at the bottom, but I normally use the ESC key.)

(PS: Similarly, in the Face Manager window, the close button moves to the top when there are no faces. Again, I mostly use the ESC key, but its funny that I have to chase the button when I've eliminated all the unconfirmed faces.)

3. Batch facial recognition seems really fast, but the results are different than pressing F6 in the file viewer (F6 in the viewer does a better job.) Does the F6 in the viewer us the "Original Size" option or "Smaller Faces" option for face recognition? I haven't tried either of these options yet.

4. It would be fun if the People View had a bit more information (number of trained faces!) and maybe the ability to sort? I have some people that are over trained (100 faces?) If the bottom line of the detailed tile said "2500 files. 30 unconfirmed. 12 trained.", then it would be easier to find the problem people. Sorting is less important because I found the "only show unconfirmed" button at the top, but it would be nice to sort by number of files or number of trained faces.

[Mario]

3. Batch facial recognition seems really fast, but the results are different than pressing F6 in the file viewer

The Viewer uses the "Optimize for small faces" always. See Special Cases

2) If you need to rerun face detection on an image which already has face annotations, do Ctrl+A, Del, F6

(PS: A little goofy that the close box on the Person Selector is at the bottom, but I normally use the ESC key.)

This was established during the beta test phase of IMatch 2020. It works best that way for users (same position as the Cancel key in regular dialog boxes).

I did not want to pollute the Person View with too many technical data.
100 trained faces is way to much. IMatch trains only up to 10 faces automatically.

[CVKen3]

The Viewer uses the "Optimize for small faces" always. See Special Cases

Ah, missed that in the help file! There's so much info there it often takes me a couple of passes to get it.

2) If you need to rerun face detection on an image which already has face annotations, do Ctrl+A, Del, F6

Not a great option. "F6, down arrow, ENTER" is probably more efficient. Just wondering if there was a way to do this in a single keystroke.

Maybe my feature request is really for key mapping?

Thanks for the quick answer!

[CVKen3]

1. If I add a manual face annotation (draw a box), it either comes up with a name or a question mark.

OK, I discovered that you can just draw the box and press F2 without any other steps, which is almost exactly what I want. It would be perfect if hitting the ENTER key with a selected face was the same as F2. Then I could draw the box and hit ENTER.

I suspect this is because I keep my mouse on the left and my right hand on the arrow keys on my keyboard! Damn those non-standard users!

Thanks for your time.