MixSTE: Seq2seq Mixed Spatio-Temporal Encoder for 3D Human Pose Estimation in Video | IEEE Conference Publication | IEEE Xplore