Face recognition - avoid messing up the data

[Tom Cunningham]

I'm with Ian. I guess I really don't understand face recognition, or how it is supposed to operate in the digiKam workflow. Like Ian, I will tag a person even if their face isn't showing (back of head) or is partially obscured (I usually have to have at least some of the head showing, however). I do all of this manually. If I remember correctly in Picasa it would attempt to identify a face and you could accept or change it on the fly. Am I correct in saying that this is more or less a batch operation in digiKam that can be performed at the album or sub-album level? So in the workflow one would import pictures and run this immediately, then check if digiKam correctly identified any faces, and manually fix those that were missed or mis-identified? I think I could live with that. What I couldn’t live with is, if in order to populate a face recognition database I ran the software against an album that already contained face tags and digiKam somehow removed/altered/duplicated the tags. That would be mildly catastrophic. Is there already a Bugzilla entry addressing these issues?

Thanks,

Tom


On Mon, 23 Oct 2023 22:42:22 +0100, Ian McCarthy <idm.kde at cronkshaw.com> wrote:

> I have a major problem with face tagging. Since the early days of Digikam I have been using face tag to mark people. I can recognise the back of people, and use lots of other methods, such as the clothes they are wearing etc. So I have a large number of images 50K plus so marked.
> I knew this problem would be coming towards me, and that software would appear that would help me, excellent. So I need a method of grading the quality of the face tags, such that the software uses the ones that it can cope with, and I can continue to mark people in images, including the backs of there heads etc.

> This I suspect would help a lot of people,and is a facility request.