[digiKam-users] Face recognition with existing faces on 7.0.0

[Simon Frei]

I just finally found time to build and try the new face management and stumbled over the following. I already have a big amount of tagged faces from 6.x. Now I ran detect faces on all images, that already contain a face. Next I wanted to run face detection on the same set of images (lots of unrecognized faces there). However I then get the error message "Face recognition is aborted, because there are no identities to recognize. Please add new identities.". I would have expected, that it automatically uses existing face tags for "training". Did I miss an option to do that? Or is it not doable automatically, but maybe requires some one-time transition when upgrading? Because manually retagging a few images for every person in my db would be quite a bit of work.
Anyway thanks to Maik and Gilles (from the git log it looks like you two still do all the heavy lifting ;) ) and all GSoC students and other contributors for your work on digikam!

Cheers,
Simon

[woenx]

I noticed the same thing. I have several thousand tagged faces already from
previous digikam versions, and the recognizer only found faces for those
that I manually tagged in 7.0, so I opened this bug:
https://bugs.kde.org/show_bug.cgi?id=415566



--
Sent from: http://digikam.1695700.n4.nabble.com/digikam-users-f1735189.html

[Simon Frei]

On 29/12/2019 15:11, woenx wrote:
> I noticed the same thing. I have several thousand tagged faces already from
> previous digikam versions, and the recognizer only found faces for those
> that I manually tagged in 7.0, so I opened this bug:
> https://bugs.kde.org/show_bug.cgi?id=415566

I just compiled master and it seems to be fixed there (should have done
that in the first place before asking, sorry) - so just test the latest
daily appimage (or tomorrows, don't know when the change happend) or
wait for beta2 ;)

And there also seems to be a tone of issues active in bugzilla about
face detection - I didn't expect that.

[Gilles Caulier-4]

In fact, if you look in AppImage date, it a little bit outdated in daily build directory. I will fix it later tomorrow, when i go back at home.

Have a happy new year.

Gilles

Le dim. 29 déc. 2019 à 15:10, Simon Frei <[hidden email]> a écrit :

On 29/12/2019 15:11, woenx wrote:
> I noticed the same thing. I have several thousand tagged faces already from
> previous digikam versions, and the recognizer only found faces for those
> that I manually tagged in 7.0, so I opened this bug:
> https://bugs.kde.org/show_bug.cgi?id=415566

I just compiled master and it seems to be fixed there (should have done
that in the first place before asking, sorry) - so just test the latest
daily appimage (or tomorrows, don't know when the change happend) or
wait for beta2 ;)

And there also seems to be a tone of issues active in bugzilla about
face detection - I didn't expect that.

[woenx]

Hey. I wanted to re-train my database with my existing faces, but now I
realize that the option to "Delete training data and rebuild" is no longer
there. What is the alternative? Is the cleanup performed automatically now?

I am using the last beta available
(digikam-7.0.0-beta2-20200105T130918-x86-64.appimage)



--
Sent from: http://digikam.1695700.n4.nabble.com/digikam-users-f1735189.html

[Gilles Caulier-4]

option is now moved in maintenance tool.

Best

Gilles

Le lun. 6 janv. 2020 à 01:28, woenx <[hidden email]> a écrit :

Hey. I wanted to re-train my database with my existing faces, but now I
realize that the option to "Delete training data and rebuild" is no longer
there. What is the alternative? Is the cleanup performed automatically now?

I am using the last beta available
(digikam-7.0.0-beta2-20200105T130918-x86-64.appimage)



--
Sent from: http://digikam.1695700.n4.nabble.com/digikam-users-f1735189.html

[woenx]

Oh, ok. It makes sense. Thanks!



--
Sent from: http://digikam.1695700.n4.nabble.com/digikam-users-f1735189.html