Way to batch ignore/disregard confirmed faces while keeping the corresponding people tag to the image?

Software digiKam
1

Hi, I am using digikam 8.7.0.
I review my collection of confirmed faces to find some of them are of low quality, e.g.:
1.blur face
2.large part of face containing unique features blocked by obstacles
3.faces overlapped
4. the person looking away

The digikam indeed does a good job picking the right guy. However, when there are more than enough good quality confirmed face samples, I’d like to remove these subpar faces from database for future face recognition guidance, since I think they’d have introduced more false positive suggestions.

It’s quite easy to do with unconfirmed faces. As suggested by the document, I right click on the unconfirmed face, assign people tag to the whole image, then press “ignore” button under the said face. And this can be done to a selected group of faces.

But with confirmed faces, I can not find a quick way to do it. All I can do is select them, then remove the people tag, hopefully they can come up as suggested unconfirmed faces during the next iteration.

Is there a way I can batch ignore/disregard these confirmed subpar faces, while keeping the people tag attached to the photo?

1 Like
2

I read a thread on th digikam user mailing list about this recently (-ish, i’d estimate few months ago), with the gist being: You don’t have to.
I don’t have time to search for it right now, but should be reasonably easy to find. The reason is that digikam itself already does a selection based on “quality”.
As for the concrete question how to do that regardless, I unfortunately don’t know.

3

As I‘ve done some research and always use the latest release of digikam (8.8.0 as of now), digikam has introduced Face Image Quality Assessment (FIQA) since 8.6.0.

From my experiences, as those suggested faces provided by digikam, it does disregard blur,low res or high iso face samples, however it does not handle “2.large part of face containing unique features blocked by obstacles 3.faces overlapped 4. the person looking away” if those face are quite clear, AKA, good enough from FIQA perspective.

So it somewhat helps, users still have to take extra care on certain face should be directly confirmed though.

In short, it does not solve my original problem.