Bug 466564 - Ability to rescan face regions to get better results
Summary: Ability to rescan face regions to get better results
Status: RESOLVED FIXED
Alias: None
Product: digikam
Classification: Applications
Component: Faces-Detection (show other bugs)
Version: 8.0.0
Platform: Mint (Ubuntu based) Linux
: NOR wishlist
Target Milestone: ---
Assignee: Digikam Developers
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2023-02-27 22:37 UTC by Petr Schonmann
Modified: 2023-03-07 03:53 UTC (History)
1 user (show)

See Also:
Latest Commit:
Version Fixed In: 8.0.0
Sentry Crash Report:


Attachments
01_before_rescan (305.00 KB, image/jpeg)
2023-02-27 22:37 UTC, Petr Schonmann
Details
02_after_rescan (305.21 KB, image/jpeg)
2023-02-27 22:38 UTC, Petr Schonmann
Details

Note You need to log in before you can comment on or make changes to this bug.
Description Petr Schonmann 2023-02-27 22:37:59 UTC
Created attachment 156804 [details]
01_before_rescan

SUMMARY
***
I have lots of pohotos tagged in  Lightroom, seems Digikam have better face detection. Should be fine to rescan old face regions to get better results to recognize faces.
***


STEPS TO REPRODUCE
1. Open the photo with "smaller" region ( 01_before_rescan )
2. Face not match the bigger one ( 02_after_rescan )

OBSERVED RESULT

Detected face region in digikam is better because match bigger face region and i think i get better results with these faces region in second stage - recognize faces ( more information :D to compute )

EXPECTED RESULT

I dont know is its possible, but would be fine to get "rescan" these face regions and fix them. Tags must be preserved in this process and batch processing in maintenance window would be great too

Lets discuss :D !!! Im still learning with digikam

SOFTWARE/OS VERSIONS
Linux/KDE Plasma:  Mint 21.1
(available in About System)
KDE Plasma Version: 
KDE Frameworks Version: 
Qt Version: 

ADDITIONAL INFORMATION
Compiled from source
Build date: 27.02.23 19:57 (target: debug)
Revision: 82bba01fd6a2fcd795dee80267b479563fabf88b
Branch: master
Comment 1 Petr Schonmann 2023-02-27 22:38:19 UTC
Created attachment 156805 [details]
02_after_rescan
Comment 2 caulier.gilles 2023-02-28 01:21:15 UTC
Try "Scan Again And Merge Result" :

https://docs.digikam.org/en/main_window/people_view.html#face-detection

Also using "Yolo v3" model, this will give a better detections.

Best 

Gilles Caulier
Comment 3 Petr Schonmann 2023-03-04 15:54:23 UTC
(In reply to caulier.gilles from comment #2)
> Try "Scan Again And Merge Result" :
> 
> https://docs.digikam.org/en/main_window/people_view.html#face-detection
> 
> Also using "Yolo v3" model, this will give a better detections.
> 
> Best 
> 
> Gilles Caulier

Scan Again and Merge results doesnt work. Same results when rescan folder.
When i delete all faces in photo and detect faces again, it works perfect. Faces region are detected better, but must tag face again :)
Comment 4 Maik Qualmann 2023-03-06 21:21:39 UTC
Git commit 2f573635a9e8b1e9ba44c67af8e5dbfc80f68ec5 by Maik Qualmann.
Committed on 06/03/2023 at 21:20.
Pushed by mqualmann into branch 'master'.

add option to clear all previous results and rescan for faces
FIXED-IN: 8.0.0

M  +1    -1    NEWS
M  +2    -1    core/utilities/facemanagement/database/facescansettings.h
M  +4    -0    core/utilities/facemanagement/threads/facepipeline.h
M  +4    -3    core/utilities/facemanagement/widgets/facescanwidget.cpp
M  +7    -1    core/utilities/facemanagement/workers/databasewriter.cpp
M  +5    -0    core/utilities/maintenance/facesdetector.cpp
M  +4    -3    core/utilities/maintenance/maintenancedlg.cpp

https://invent.kde.org/graphics/digikam/commit/2f573635a9e8b1e9ba44c67af8e5dbfc80f68ec5
Comment 5 caulier.gilles 2023-03-07 03:53:03 UTC
Git commit 81ef9f69a9748d4448599d437a5204679bf24062 by Gilles Caulier.
Committed on 07/03/2023 at 03:50.
Pushed by cgilles into branch 'master'.

digiKam doc: add new faces management option to clear all previous faces in database and rescan

M  +2    -0    main_window/people_view.rst

https://invent.kde.org/documentation/digikam-doc/commit/81ef9f69a9748d4448599d437a5204679bf24062