Bug 316898

Summary: Manage face area position as persistent data in database accordingly with image manipulations
Product: [Applications] digikam Reporter: Axel Krebs <axel.krebs>
Component: Database-FacesAssignee: Digikam Developers <digikam-bugs-null>
Status: RESOLVED DUPLICATE    
Severity: normal CC: bugskdeorg.20.dimon3000, caulier.gilles, mario.frank, mike, spam-receiver
Priority: NOR    
Version: 3.0.0   
Target Milestone: ---   
Platform: Ubuntu   
OS: Linux   
Latest Commit: Version Fixed In:
Sentry Crash Report:

Description Axel Krebs 2013-03-17 11:01:34 UTC
when detecting _several_ faces on _one_ pic, the position of _each_ face should be associated and stored ("somehow") persistent in absolute data. 

Reason: 
- if working on a pic, the pic can be divided. Up to know, the faces info stay with both pics- which is wrong, potentially. 
- if showing _one_ face of several, one still sees all faces names, even in this extreme case wher only ony face is visible in the observed area. 



Reproducible: Always

Steps to Reproduce:
1. run face recognition
2. observe the above mentioned appearences
3. 


Expected Results:  
every face should have its own (persistent) position information
Comment 1 caulier.gilles 2014-09-05 09:39:50 UTC
Marcel,

Face area positions are not already stored as persistent in database ?

Gilles
Comment 2 Marcel Wiesweg 2014-09-05 18:55:14 UTC
Yes, it is always the tag (the person) and the area on the photo
Comment 3 Mario Frank 2017-01-17 10:00:53 UTC
(In reply to Marcel Wiesweg from comment #2)
> Yes, it is always the tag (the person) and the area on the photo

Marcel is right. The face area is persistent in database. For every face,
there is an entry with image id, face tag id and the tag region/autodetected face as absolute coordinates (x,y, width and heigth) to the specific image.

Is this still valid? What behaviour is exactly expected?
Comment 4 caulier.gilles 2017-01-17 10:43:41 UTC
Mario,

Imagine a photo with a face already registered as face tag.

Open the image, and crop it. The face area is changed. The expected behavior is to recompute face area accordingly with manipulations. The can be a perspective adjustment, a flip, a rotation, a crop, etc.

And this must be true also with BQM tools.

So this can be a complex patch to do.

Gilles
Comment 5 bugskdeorg.20.dimon3000 2018-11-13 13:33:51 UTC
Hello.
Any estimates on this?
Comment 6 caulier.gilles 2018-11-13 13:57:19 UTC
Hi,

Currently, all internal test tools for FaceEngine need to be ported as real unit-tests. After this stage, it will be more easy to check broken or missing features.

I already ported all MetaEngine as unit-tests. This take few weeks of work and permit to identify and fix serious problems. The same must be done for FaceEngine.

In all cases, i want to complete FaceEngine unit-tests before 6.0.0 final release.

Best

Gilles Caulier
Comment 7 spam-receiver 2019-07-27 12:28:30 UTC
(In reply to caulier.gilles from comment #6)
> Hi,
> 
> Currently, all internal test tools for FaceEngine need to be ported as real
> unit-tests. After this stage, it will be more easy to check broken or
> missing features.
> 
> I already ported all MetaEngine as unit-tests. This take few weeks of work
> and permit to identify and fix serious problems. The same must be done for
> FaceEngine.
> 
> In all cases, i want to complete FaceEngine unit-tests before 6.0.0 final
> release.
> 
> Best
> 
> Gilles Caulier

Have the test tools been separated from productive code?
What's he status of the face recognition & management feature?
Comment 8 caulier.gilles 2023-05-07 08:42:00 UTC
Maik,

I suppose that entry depend of other bug 286529 where the face area must be updated in metadata if transform tools are processed on images. Right ?

Gilles
Comment 9 caulier.gilles 2023-10-12 14:41:53 UTC
Maik, did you seen my last comment ?
Comment 10 caulier.gilles 2024-10-08 05:37:10 UTC

*** This bug has been marked as a duplicate of bug 286529 ***