SRNet: Improving Generalization in 3D Human Pose Estimation with a Split-and-Recombine Approach | Synapse