Look, Cast and Mold: Learning 3D Shape Manifold from Single-view Synthetic Data