This is the implementation of the approach described in the paper: Dario Pavllo, Christoph Feichtenhofer, David Grangier, and Michael Auli. 3D human pose estimation in video with temporal convolutions ...