Pixel-Aligned Volumetric Avatars

Figure 1 : We present a novel approach for the prediction of volumetric avatars of human heads from a small number of example views. Our model enables view synthesis for unseen identities and is able to generate faithful facial expressions.