The Face Recognition Prize Challenge (FRPC) 2017 seeks to identify which algorithms are most accurate at recognizing faces in images collected in an unconstrained manner. It also seeks to assess the limits of recognition and to understand when algorithms fail. The term unconstrained refers to images for which there is weak control of the photographic conditions and the behavior of the human subject. Unconstrained face images are characterized by: variable head pose relative to the position of the camera; variable head position in the scene; variable head size; variable and non-uniform ambient illumination; varying degrees of facial expression, with and without open mouth; occlusion of the face by hair, arms or other objects.