Why doesn't emotion classification work as specified?

asked 2015-07-05 01:06:59 -0500

mayooran gravatar image

I went through the Gender classification tutorial from openCV, http://docs.opencv.org/modules/contri.... I need to do emotion classification for which they say, "If you want to do emotion classification instead of gender classification, all you need to do is to update is your training data and the configuration you pass to the demo" . I used a dataset with 16 images each for happy and sad emotions. Happy dataset includes faces of happy men,women,kids and elders. Similarly sad dataset as well. But when I tested this on 4 images all 4 went wrong. What other changes should I do apart from the dataset as said in the link by, "configurations you pass to the demo". What configurations should I change from the gender classification example facerec_fisherfaces.cpp? I am new to OpenCV. Please help.

edit retag flag offensive close merge delete



"includes faces of happy men,women,kids and elders." - in that case, you need far more data, try with 200 per class, 16 are definitely not enough. also take care to proerly crop and align the face images.

berak gravatar imageberak ( 2015-07-05 01:34:51 -0500 )edit

They have been cropped using the python script available in the same link. So will adding more images to the model help? don't I need to change any configurations? And I am using the FisherFaces algorithm. Is that okay?

mayooran gravatar imagemayooran ( 2015-07-05 05:42:51 -0500 )edit

ther isn't much to tweak in your situation, i guess.

berak gravatar imageberak ( 2015-07-05 07:04:07 -0500 )edit