Pose Embeddings: A Deep Architecture for Learning to Match Human Poses

   Abstract