Learning to discriminate face views