Bug 471335 - Tag a detected face but not use it for training
Summary: Tag a detected face but not use it for training
Status: RESOLVED FIXED
Alias: None
Product: digikam
Classification: Applications
Component: Faces-Workflow (other bugs)
Version First Reported In: 7.9.0
Platform: Debian unstable Linux
: NOR wishlist
Target Milestone: ---
Assignee: Digikam Developers
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2023-06-22 12:50 UTC by Marcos Dione
Modified: 2025-08-20 07:53 UTC (History)
5 users (show)

See Also:
Latest Commit:
Version Fixed/Implemented In: 8.7.0
Sentry Crash Report:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Marcos Dione 2023-06-22 12:50:31 UTC
The face detection algorithm is very good, to the point where it detects faces out of focus, in really bad angles, etc.

While using the People view to tag faces, we can tag these 'bad' faces. It would be nice if we could tell digikam not to use them for training the face recognition tool. I have the impression that when doing so, the algo starts confusing all the out of 'bad' faces, so the algo's hit rate plummets.

Maybe related to https://bugs.kde.org/show_bug.cgi?id=425263
Comment 1 Marcos Dione 2023-06-22 12:56:03 UTC
(In reply to Marcos Dione from comment #0)
> The face detection algorithm is very good, to the point where it detects
> faces out of focus, in really bad angles, etc.

Other reasons for 'bad' photos:

* Partially occluded
* Bad lightning
Comment 2 caulier.gilles 2023-06-22 13:42:40 UTC
Hi Tran,

Another file from your Face Management improvements Google Summer of Code 2023 project 

Best

Gilles
Comment 3 Chris 2023-10-17 08:51:19 UTC
Thumbs up for this suggestion
Comment 4 Maik Qualmann 2024-06-02 08:45:56 UTC
*** Bug 487911 has been marked as a duplicate of this bug. ***
Comment 5 Michael Miller 2025-04-27 02:38:31 UTC
Fixed with the addition of FIQA
Comment 6 Marcos Dione 2025-08-20 07:52:40 UTC
Hi. This was fixed in Apr27, and there was a release on Jun30. Would this fix be in that release?
Comment 7 caulier.gilles 2025-08-20 07:53:37 UTC
yes sure