Video-based face recognition using the intra/extra-personal difference dictionary

Face recognition in unconstrained videos is challenging due to large variations in pose, illumination, expression etc. We address the problem from two different aspects: To handle pose variations, we learn a Structural-SVM based detector which can simultaneously localize face fiducial points and estimate the face pose. By adopting a different optimization criterion from existing algorithms, we are able to improve localization accuracy. To model other face variations, we use intra-personal/extra-personal dictionaries. The proposed framework is advantageous in terms of both accuracy and scalability. We demonstrate through experiments that our algorithm achieves state-of-arts performance on challenging public databases, even when the training data come from a different database.