Learning Shape Priors for Single-View 3D Completion and Reconstruction