Bug 384614 - Digikam redeye does not work on some images
Summary: Digikam redeye does not work on some images
Status: RESOLVED FIXED
Alias: None
Product: digikam
Classification: Applications
Component: Faces-RedEyes (show other bugs)
Version: 5.7.0
Platform: Other Linux
: NOR normal
Target Milestone: ---
Assignee: Digikam Developers
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2017-09-12 09:35 UTC by humufr
Modified: 2023-10-19 06:30 UTC (History)
3 users (show)

See Also:
Latest Commit:
Version Fixed In: 7.2.0


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description humufr 2017-09-12 09:35:40 UTC
I took some picture this summer and one of the subject does have very clear eyes and so he is subject to redeye. I try to use digikam tool to remove them but there are absolutely no visible difference in the before and after image. I try to change the Red level but it doesn't have any effect and I am not sure what is the usage of the channel and curve.

I am probably not using the tool properly but I am probably not the only one in this case.

I try to use the redeye removal tool from The Gimp and it worked correctly.

I can send the picture but I will prefer to not put them on a public website. I am sorry but there are privacy issue here.

Thanks for digikam, it is a great tool.
Comment 1 Maik Qualmann 2017-09-12 10:35:45 UTC
You can send me a test image to my private email address.

Maik
Comment 2 Maik Qualmann 2017-09-12 17:56:20 UTC
Thanks for the test image. The openCV face detection does not recognize the face in the image, therefore, also no recognition of the red eyes. The face is slightly in profile. Just take a look...

Maik
Comment 3 humufr 2017-09-13 08:15:36 UTC
Perhaps the tool is a little bit too sophisticated and another option where selecting the eye and detecting an ellipse with red-ish color can be a good addition but perhaps you will be able to improve the automatic detection which will be very helpful for most of people.

Thanks to look at the problem.
Comment 4 caulier.gilles 2017-09-13 08:20:52 UTC
A manual patch will not be able to run in batch with all automatized workflow.

The algorithm to detect the face and red eyes need to be improved. We have a student working on face recognition while this summer who make great Face Management improvements. He is a mathematician, so certainly he can take a look if time permit.

I will CC him.

Gilles Caulier
Comment 5 caulier.gilles 2017-09-13 08:49:49 UTC
Lui,

I CC you about this entry, to see if you can take a look in the algorithm for some face detections improvements (including eyes).

Thanks in advance

Gilles Caulier
Comment 6 caulier.gilles 2017-09-13 08:50:54 UTC
Hulufr,

Can you share the problematic image somewhere in cloud for future investiguations ?

Thanks in advance.

Gilles Caulier
Comment 7 Maik Qualmann 2021-02-13 16:50:13 UTC
Git commit 99efff24b22569e258bc712c3befef50684f20c4 by Maik Qualmann.
Committed on 13/02/2021 at 16:48.
Pushed by mqualmann into branch 'master'.

enable Yolo v3 face detection for the red eye tool
FIXED-IN: 7.2.0

M  +2    -1    NEWS
M  +4    -0    core/libs/dimg/filters/redeye/redeyecorrectionfilter.cpp

https://invent.kde.org/graphics/digikam/commit/99efff24b22569e258bc712c3befef50684f20c4
Comment 8 Maik Qualmann 2023-10-19 06:30:37 UTC
Git commit d198300bf06951b9a03951af44912525f28a5165 by Maik Qualmann.
Committed on 19/10/2023 at 08:28.
Pushed by mqualmann into branch 'master'.

use YoloV3 face detection again for better red eye detection
Adjust the size of detected faces for problematic
images that didn't work with YoloV3.
Related: bug 463350

M  +12   -10   core/libs/dimg/filters/redeye/redeyecorrectionfilter.cpp

https://invent.kde.org/graphics/digikam/-/commit/d198300bf06951b9a03951af44912525f28a5165