In this work, we investigate the problem of creating high-fidelity 3D content from only a single image.
In contrast to the traditional avatar creation pipeline which is a costly process, contemporary generative approaches directly learn the data distribution from photographs.
Moreover, for effective training, we consider difficulty-based sampling strategy to encourage the network to pay more attention to some partial point clouds with fewer geometric information.
An asymmetric keypoint locator, including an unsupervised multi-scale keypoint detector and a complete keypoint generator, is proposed for localizing aligned keypoints from complete and partial point clouds.
Most existing expression manipulation methods resort to discrete expression labels, which mainly edit global expressions and ignore the manipulation of fine details.
Instead of using an intermediate estimated guidance, we propose to explicitly transfer facial expression by directly mapping two unpaired input images to two synthesized images with swapped expressions.