Global-to-Local Modeling for Video-Based 3D Human Pose and Shape Estimation | Read Paper on Bytez