Ignored Faces

[Kartik Ramesh]

Thanks for clarifying, Gilles.
Yes, as Trung mentioned, the purpose of the red round button is not to
Ignore Faces. Trung, your observations are correct except for :

> Clicking on the "*-*" button turns an unconfirmed face to ignored face
>
 Ignored -> Unknown. As we are rejecting the face suggestion.

Currently the functionality of the buttons is as follows:
1. Red Round Button (top right) : delete Face from database, regardless of
what type of face it is. The red round button currently shows the tool-tip
"If this is not a face, click to reject it", maybe I could modify it to
emphasize that the face will be deleted.
2. (-) Button (bottom right) : move face to Unknown, except if face is
already Unknown. In that case, move it to Ignored. This button is disabled
for Ignored.
3. Green checkbox (bottom left) : Confirms the Face. For Unconfirmed Faces
(i.e. suggestions), this corresponds to marking as Confirmed. For Ignored
Faces, however, this leads to a marking of the face as Unknown. I preferred
this because it seemed intuitive, as if the user is saying "Ok, I'm
allowing this face to be detected now". I've displayed the tool-tip
accordingly.

Also another thing, earlier Maik had mentioned that the Ignored Face
shouldn't be storing the tag region in the database, as the user doesn't
care for the region anymore. I liked the idea at first, and implemented the
tagRegion for Ignored tag as a NULL string but I ran into two issues :
1. How to detect overlaps of Face Region between faces being recognized
(during a recognition process) and the faces marked as Ignored.
2. How to show the face in the Ignored View (when we selected Ignored tag
from People Sidebar), as the default implementation to depict faces without
a region is to show the entire image.

I could have worked around with a dummy placeholder (as Maik suggested),
but I thought to keep the implementation simple. Anyhow, in the current
implementation, the image is not being associated with the Ignored face,
i.e. if you mark a face as Ignored, and then open that image (in Digikam
album view), then the ignored face will not be displayed. So it works fine.

Kartik

On Sun, May 17, 2020 at 5:56 PM Gilles Caulier <caulier.gilles at gmail.com>
wrote:

> This one probably :
>
> https://imgur.com/m5lftgN
>
> Gilles Caulier
>
> Le dim. 17 mai 2020 à 14:17, Kartik Ramesh <kartikx2000 at gmail.com> a
> écrit :
> >
> > Thank you for your responses :)
> > I will share my thought on the implementation, but before that could
> someone please clarify what is the "round red button" is being referred to?
> Maybe I'm using a different theme, so please clarify, because I don't want
> to start explaining without completely understanding the doubt.
> >
> > Thanks
> > Kartik
> >
> > On Sun, May 17, 2020 at 4:17 PM Maik Qualmann <metzpinguin at gmail.com>
> wrote:
> >>
> >> Hmm, in my opinion the round red button should ignore the face and no
> longer
> >> display it in future face detections. The "-" button should make an
> >> unconfirmed face an unknown face again.
> >>
> >> Maik
> >>
> >> Am Sonntag, 17. Mai 2020, 12:07:24 CEST schrieb Thanh Trung Dinh:
> >> > The implementation is really good, well done Kartik!
> >> > Besides, talking about the red button, I think its functionality is
> not for
> >> > turning unknown faces to ignored faces. Clicking on the "*-*" button
> turns
> >> > an unconfirmed face to ignored face, but clicking red button will
> remove
> >> > face tag completely I suppose. Then, when you run face detection
> again with
> >> > option "skip already scanned images", those faces (rejected by the red
> >> > button) will not appear, but when you run again with option "scan
> again and
> >> > merge results", those faces will be rescanned and appear again if
> there are
> >> > faces detected. For me, the functionality of the red button is quite
> >> > confused. Correct me if my observation above is wrong, but I think
> that the
> >> > red button should be hidden on the overlay. What do you think?
> >> >
> >> >
> >> >
> >> > On Sun, May 17, 2020 at 11:07 AM Maik Qualmann <metzpinguin at gmail.com
> >
> >> >
> >> > wrote:
> >> > > I like the implementation very much, good job. But if I ignore a
> face (the
> >> > > red
> >> > > round overlay button), it is found again. I did the manual
> detection on an
> >> > > image. When I look to the "ImageTagProperty" table the region was
> deleted
> >> > > and
> >> > > didn't get the "Ignored" Tag ID. So it will be found again.
> >> > >
> >> > > Maik
> >> > >
> >> > > Am Sonntag, 17. Mai 2020, 08:59:07 CEST schrieb Kartik Ramesh:
> >> > > > Hello all, I've managed to implement the feature for Ignored
> Faces. I've
> >> > > > pushed to the digikam repo. please let me know if it's
> appropriate.
> >> > > >
> >> > > > The basic idea is as follows :
> >> > > > 1. Pressing Reject on Unknown Faces, should mark the face as
> Ignored.
> >> > >
> >> > > This
> >> > >
> >> > > > will create a new tag, and move the face to it.
> >> > > > 2. Running a Face Detection now, will not re-detect the Ignored
> Face
> >> > >
> >> > > (even
> >> > >
> >> > > > if Scan again and Merge is selected). Facial Recognition will not
> try to
> >> > > > recognize the ignored faces, and will never suggest names for an
> ignored
> >> > > > face.
> >> > > > 3. I have re-designed overlay for Ignored Faces in particular.
> Pressing
> >> > >
> >> > > the
> >> > >
> >> > > > confirm button, on an Ignored Face will un-mark the face as
> Ignored, and
> >> > > > move it to Unknown.
> >> > > >
> >> > > > Changes made to directories:
> >> > > > core/app
> >> > > > core/utilities
> >> > > > core/libs
> >> > > >
> >> > > > To do:
> >> > > > 1. The people sidebar should display Ignored Faces at the bottom,
> as it
> >> > > > isn't of priority to the user.
> >> > > > 2. Add Ignored Property to the metadata of the image, as discussed
> >> > >
> >> > > earlier.
> >> > >
> >> > > > Regards
> >> > > > Kartik
> >>
> >>
> >>
> >>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/digikam-devel/attachments/20200517/6b0a558f/attachment-0001.htm>