Temporal Invariant Facial Recognition

asked 2014-09-23 12:58:26 -0500

mh00h gravatar image

updated 2014-09-23 13:24:55 -0500

berak gravatar image


Is there a way to do facial recognition that is age invariant (temporal invariant)? Relatedly, are eigenfaces (or fischerfaces) age invariant given enough training images at different ages?

Thank you.

edit retag flag offensive close merge delete


(not really an answer, just food for thought:)

if you want to detect images from persons at different ages, ofc. you will have to add photos from that age.

unfortunately, doing so will increase the inner-class-variance, and downgrade your overall recognition.

you'll probably get better results with lbph than with fisher or eigen in this situation.

berak gravatar imageberak ( 2014-09-24 04:36:35 -0500 )edit