We propose DeformingNet, an effective 3D point cloud completion network. Unlike existing methods that complete partial point cloud by directly learning the morphing function from 2D grids to 3D shapes ...