During last year's coding sprint, I investigated existing formats to
highlight regions. You can find the document here:
http://websvn.kde.org/trunk/extragear/graphics/digikam/project/ImageAnnotation.odt
Feel free to add to it once you get SVN access ;-)
As for resizing of the images: This would risk making faces too small if
there are lots of people in the image, wouldn't it? Consider a group
photo of people, which may be 3000x2000 pixels, but the individual faces
are only 100x100 pixel. But that is probably a detail that can be found
out later.
Michael
Hi Michael,
I'm applying for an KDE svn account soon. SVN should be unblocked in my campus soon, too.
Thanks for the document.
True, if the face size is 100*100 pixels, then resizing and detecting could be a problem. Currently libface correctly detects faces of size 50*50 pixels, faces smaller than that don't get detected nicely. So that IS a problem. But once we detect the tiny faces, we can resize the region box for the image of the original scale, and then clip the face from the original image for training. So here, it won't damage training by feeding bad quality faces, it will only damage detection.
Thanks,
--
Aditya Bhatt
_______________________________________________
Digikam-devel mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-devel
Locked
