Inferring 3D shape and appearance from natural images is a fundamental challenge in computer vision. Despite recent progress using deep learning methods, a key limitation is the availability of annotated training data, as ...