Make It Easier: An Empirical Simplification of a Deep 3D Segmentation Network for Human Body Parts