Fine-grained pose prediction, normalization, and recognition
2015·Arxiv