A semantic model for video based face recognition.