Relightable and Animatable Neural Avatar from Sparse-View Video

Figure 1 . Reconstructing relightable and animatable neural avatar from sparse-view (or monocular) video. Our method takes only a sparse-view (or monocular) video as input and reconstructs a relightable and animatable neural avatar under unknown illumination, which can then be relit with arbitrary environment lights and animated with arbitrary motion sequences. Note that our method successfully captures the shininess of the skin and pants as well as the specular highlights on the t-shirt's plastisol printings.