Deep 3D Portrait From a Single Image

In this paper, we present a learning-based approach for recovering the 3D geometry of human head from a single portrait image. Our method is learned in an unsupervised manner without any ground-truth 3D data. We represent the head geometry with a parametric 3D face model together with a depth map for other head regions including hair and ear. A two-step geometry learning scheme is proposed to learn 3D head reconstruction from in-the-wild face images, where we first learn face shape on single images using selfreconstruction and then learn hair and ear geometry using pairs of images in a stereo-matching fashion. The second step is based on the output of the first to not only improve the accuracy but also ensure the consistency of overall head geometry. We evaluate the accuracy of our method both in 3D and with pose manipulation tasks on 2D images. We alter pose based on the recovered geometry and apply a refinement network trained with adversarial learning to ameliorate the reprojected images and translate them to the real image domain. Extensive evaluations and comparison with previous methods show that our new method can produce high-fidelity 3D head geometry and head pose manipulation results. * This work was done when S. Xu was an intern at MSRA. tates substantial image content regeneration in the head region and beyond. Promising results have been shown for face rotation [58, 3, 27] with generative adversarial nets (GAN), but generating the whole head region with new poses is still far from being solved. One reason could be implicitly learning the complex 3D geometry of a large variety of hair styles and interpret them onto 2D pixel grid is still prohibitively challenging for GANs.