HI.
I am using face_recognition and face_detection to tag my pictures but often they cannot detect any face if it is small and unclear in the picture.
The same face which is been extracted in a new picture, so zoomed, is easly detected.
Is there any way to improve face_recognition use in order not to lose those faces?
Thanks.
Which program is (not) doing the face rcognition?
I assume this is GitHub - ageitgey/face_recognition: The world's simplest facial recognition api for Python and the command line, as used by darktable lua documentation - face_recognition. I’m not sure if @Zener is using it with darktable or not.
Yes, it is.
Thanks for your clarification!
It is an external tool used by darktable via a script. You’d have to check the original tool.
There seems to be a tolerance / sensitivity parameter: GitHub - ageitgey/face_recognition: The world's simplest facial recognition api for Python and the command line
The tolerance can be adjusted from darktable; you may want to try different values.
Though your problem seems to be face recognition not detecting faces, while the tolerance value controls the similarity threshold for recognition (labelling/tagging).
I read tolerance / sensitivity parameter for face_recognition about how it choices the most likely face.
The issue is about detect or don’t a face in a picture from face_detection (here I can’t found parameters to set).
Anyway face_recognition too can’t find any face in some pictures, I think they share the same engine.
face_recognition (from the ageitgey repo) is the engine. The Lua script just allows you to launch it from within darktable. Neither darktable, nor that Lua script do any face recognition at all.
Yes, that’ correct.
I use them (face_detection and face_recognition) on Shell when I have the issue as subject.
Nothing to improve?
Thanks.