Identifies faces in a live video stream and projects the neural-network encoded representation for recognition into 2D space