RPAN: An End-to-End Recurrent Pose-Attention Network for Action Recognition in Videos | IEEE Conference Publication | IEEE Xplore