3D Face From X: Learning Face Shape From Diverse Sources

We present a novel method to jointly learn 3D face parametric model and reconstruction from diverse sources. Previous methods usually modeling one kind of source, such as scanned data or in-the-wild images. Although contain accurate geometric information shapes, the capture system is expensive datasets small number subjects. On other hand, images are easily obtained there large facial However, do not explicit information. In this paper, we propose unified Besides images, also utilize RGB-D captured with an iPhone X bridge gap between two Experimental results demonstrate that training more sources, can powerful model.