| Summary: | Face thumbnail generation | ||
|---|---|---|---|
| Product: | [Applications] digikam | Reporter: | matthieu04 |
| Component: | Faces-Workflow | Assignee: | Digikam Developers <digikam-bugs-null> |
| Status: | REPORTED --- | ||
| Severity: | normal | CC: | caulier.gilles, iwannaberich, metzpinguin, michael_miller |
| Priority: | NOR | ||
| Version First Reported In: | 8.6.0 | ||
| Target Milestone: | --- | ||
| Platform: | Mint (Ubuntu based) | ||
| OS: | Linux | ||
| Latest Commit: | Version Fixed/Implemented In: | ||
| Sentry Crash Report: | |||
| Attachments: |
Digikam output during face thumbnail recognition
sample image Trace digikam appimage |
||
|
Description
matthieu04
2025-05-11 20:45:54 UTC
Have you downloaded all AI models files proposed at startup ? If no you can do it using DK/Setup/Miscs/System dialog page. Can you run digiKam from a console as explained in this page ? https://www.digikam.org/contribute/#linux-host Created attachment 181223 [details]
Digikam output during face thumbnail recognition
Digikam output during face thumbnail recognition
Hello, you'll find enclosed the requested traces. Thank you This looks like broken images or something similar, a sample is needed. /media/SynoCousin_photo/[Archive]/2021 - Photos/2021 Divers/20210731_131702.jpg" : Unknown image format !!! Maik Created attachment 181225 [details]
sample image
Here it is No problem with the sample image here. Strange, if the problem occurs for you, no loader plugin can identify the image path, neither via the file extension nor, in the second attempt, via the magic bytes. Network path temporarily unavailable? Maik Ah, I see Ubuntu, but you're not using the digiKam Snap package from Ubuntu? Snap is completely useless for digiKam and uses a sandbox; only local files in the home directory can be processed. Use our AppImage instead. Maik (In reply to Maik Qualmann from comment #8) > Ah, I see Ubuntu, but you're not using the digiKam Snap package from Ubuntu? > Snap is completely useless for digiKam and uses a sandbox; only local files > in the home directory can be processed. > Use our AppImage instead. > > Maik Thanks Maik. I downloaded the attached test image and everything processed normally. All face regions were detected. No false positives we created. Cheers, Mike (In reply to Maik Qualmann from comment #8) > Ah, I see Ubuntu, but you're not using the digiKam Snap package from Ubuntu? > Snap is completely useless for digiKam and uses a sandbox; only local files > in the home directory can be processed. > Use our AppImage instead. > > Maik Hello, I downloaded the appimage, here is the trace. Around 1 face is generated per 2-3seconds. Maybe it is normal since it is the first run and that files are on a NAS? Created attachment 181228 [details]
Trace digikam appimage
It feels like digikam generates the face thumbnail only when the face is being viewed, and not beforehand, when the face was detected... Is it the expected behaviour? If yes, it can be very slow, especially because it will usually be done in parallel with classifying (while a person is being classified on WAL, you want to display the next one to start classifying). It would be more practical to be able to generate all face thumbnails at once The image loading error doesn't occur with AppImage, at least. And yes, sometimes face thumbnails are created during display for newly imported images with face regions. However, if I'm right, Michael has also integrated thumbnail creation for face detection. We can look into it again, though. We already have other bug reports on the topic of face thumbnail creation. So we can basically close this one. Maik (In reply to Maik Qualmann from comment #13) > The image loading error doesn't occur with AppImage, at least. And yes, > sometimes face thumbnails are created during display for newly imported > images with face regions. However, if I'm right, Michael has also integrated > thumbnail creation for face detection. We can look into it again, though. We > already have other bug reports on the topic of face thumbnail creation. So > we can basically close this one. > > Maik Hi Maik, You are correct. The face engine creates thumbnails when the face is first detected in an image. https://invent.kde.org/graphics/digikam/-/blob/master/core/utilities/facemanagement/pipelines/detectrecognize/facepipelinedetectrecognize.cpp#L745 The thumbnail is stored in the tumbnails-digikam DB. If this database is running from a remote server, then loading the thumbnail image will be slower. Cheers, Mike Hello, My database is running on local SSD, so loading should be fast. I guess that the face thumbnail was not properly generated upon analyzing, and digikam has to re-generate it upon displaying. I can send you new traces if needed, to confirm. Otherwise you can close, if similar face thumbnail generation report are already adressed in another bug report. |