face recognizer classes
Hi!
I wanna make a software, that classifies the facial expressions into two different classes(happy, sad). I would like to use the Face Recognizer class. How many training images offered?
What about neutral faces? It is usual to build a happy/not happy classifier or a smile/non-smile classifier...
Making a happy/sad classifier will work. What you need to do for that is use facial landmarks (will work way better than the recognizer classes), then collect for example 1000 happy and 1000 sad images, apply landmarks, than run an SVM classifier on the descriptors of those landmarks. Your descriptor should somehow model the spatial relation of the keypoints.
You can use images from GENKI-4K database to train your system http://mplab.ucsd.edu/wordpress/?page...
I read this tutorial: http://docs.opencv.org/trunk/modules/... and the author mention, that this usable for classify for facial expressions. So i wanna use this.
@dragostej It is not because of the fact that the author says it is possible that it will actually work decently... but hey go ahead and report your results I would say ...