Face recognition and alignment using support vector machines

Face recognition in the presence of pose changes remains a largely unsolved problem. Severe pose changes, resulting in dramatically different appearances, is one of the main difficulties. We present a support vector machine (SVM) based system that learns the relations between corresponding local regions of the face in different poses as well as a simple SVM based system for automatic alignment of faces in differing poses. We then present experimental results from multiple random splits of the CMU PIE Database to verify the strength of our approach.