More face tag puzzlement

[Michael Miller]

Hi Tom,
If both overlapping regions are confirmed, then the face engine shouldn’t suggest a name.  It won’t suggest a name for a confirmed or ignored face region.

I found the problem with overlap detection logic and the change has been committed. It should be in the next build.

Cheers,
Mike

On May 13, 2025, at 6:20 PM, Tom Cunningham <TWCunningham627 at outlook.com> wrote:

But if the regions overlap, and the names associated with the regions are the same, shouldn’t face recognition forego making a suggestion?

Take care,

Tom


From: Michael Miller <michael_miller at msn.com>
Sent: Tuesday, May 13, 2025 4:45 PM
To: Tom Cunningham <TWCunningham627 at outlook.com>
Cc: digikam-users at kde.org
Subject: Re: More face tag puzzlement

Hi Tom,
We increased the size of the face rectangle in 8.7.0.  It sounds like the overlap detection logic is working correctly.  I’ll take a look now.

Cheers,
Mike


On May 13, 2025, at 5:27 PM, Tom Cunningham <TWCunningham627 at outlook.com<mailto:TWCunningham627 at outlook.com>> wrote:

OK, now that I’ve painstakingly cleaned up my face tags (removed duplicates, “tree” faces, and “sidewalk” faces) and installed 8.7.0 pre-release on Windows 11, I tried another scan with detection and recognition set at 10 and Recognize Faces Only. It does a fine job of recognizing faces, and it wants me to confirm the faces it finds. However, my understanding was that if a face tag already exists for a face, it will skip that face. Instead, it simply puts another rectangle inside the existing one with a (generally correct) suggestion. If I do nothing and I am mousing over faces in a picture, the suggestion will pop up inside the existing rectangle even though the face is already identified. If I ignore the suggestion, then I’m assuming it will put an Ignored box inside the existing rectangle, which doesn’t seem quite right either.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/digikam-users/attachments/20250513/40214a45/attachment.htm>