NDF: Neural Deformable Fields for Dynamic Human Modelling

We propose Neural Deformable Fields (NDF), a new representation for dynamic human digitization from multi-view video. Recent works proposed to represent body with shared canonical neural radiance fields which links the observation space deformation estimations. However, learned is static and current design of not able large movements or detailed geometry changes. In this paper, we learn deformable field wrapped around fitted parametric model human. The NDF spatially aligned by underlying reference surface. A network then map pose dynamics NDF. can synthesize digitized performer novel views poses reasonable appearance. Experiments show that our method significantly outperforms recent synthesis methods.