Learning 3D representation of a single image is challenging considering the ambiguity, occlusion, and perspective project of an object in an image. Previous works either seek image annotation or 3D supervision to learn ...